Stephen J. Thomas, Joseph A. Zeni, David A. Winter - Winter's Biomechanics and Motor Control of Human Movement-Wiley (2022)

ISTUDY
ISTUDY
WINTER’S BIOMECHANICS AND MOTOR
CONTROL OF HUMAN MOVEMENT
ISTUDY
ISTUDY
WINTER’S BIOMECHANICS
AND MOTOR CONTROL
OF HUMAN MOVEMENT
Fifth Edition
STEPHEN J. THOMAS
Thomas Jefferson University, Philadelphia, PA, USA
JOSEPH A. ZENI
Rutgers University, Newark, NJ, USA
DAVID A. WINTER (DECEASED)

University of Waterloo, Waterloo, Ontario, Canada
ISTUDY
This fifth edition first published 2023
© 2023 John Wiley & Sons, Inc.
Edition History
John Wiley & Sons (1e, 1980, 2e, 1990, 3e, 2004, 4e, 2009)
All rights reserved. No part of this publication may be reproduced, stored in a retrieval system, or transmitted, in any form or by any
means, electronic, mechanical, photocopying, recording or otherwise, except as permitted by law. Advice on how to obtain
permission to reuse material from this title is available at http://www.wiley.com/go/permissions.
The right of Stephen J. Thomas, Joseph A. Zeni, and David A. Winter be identified as the authors of this work has been asserted in
accordance with law.
Registered Office
John Wiley & Sons, Inc., 111 River Street, Hoboken, NJ 07030, USA
Editorial Office
111 River Street, Hoboken, NJ 07030, USA
For details of our global editorial offices, customer services, and more information about Wiley products visit us at
www.wiley.com.
Wiley also publishes its books in a variety of electronic formats and by print-on-demand. Some content that appears in standard
print versions of this book may not be available in other formats.
Limit of Liability/Disclaimer of Warranty

In view of ongoing research, equipment modifications, changes in governmental regulations, and the constant flow of information
relating to the use of experimental reagents, equipment, and devices, the reader is urged to review and evaluate the information
provided in the package insert or instructions for each chemical, piece of equipment, reagent, or device for, among other things, any
changes in the instructions or indication of usage and for added warnings and precautions. While the publisher and authors have
used their best efforts in preparing this work, they make no representations or warranties with respect to the accuracy or
completeness of the contents of this work and specifically disclaim all warranties, including without limitation any implied
warranties of merchantability or fitness for a particular purpose. No warranty may be created or extended by sales representatives,
written sales materials or promotional statements for this work. The fact that an organization, website, or product is referred to in
this work as a citation and/or potential source of further information does not mean that the publisher and authors endorse the
information or services the organization, website, or product may provide or recommendations it may make. This work is sold with
the understanding that the publisher is not engaged in rendering professional services. The advice and strategies contained herein
may not be suitable for your situation. You should consult with a specialist where appropriate. Further, readers should be aware that
websites listed in this work may have changed or disappeared between when this work was written and when it is read. Neither the
publisher nor authors shall be liable for any loss of profit or any other commercial damages, including but not limited to special,
incidental, consequential, or other damages.
Library of Congress Cataloging-in-Publication Data

Names: Thomas, Stephen J., 1980- author. | Zeni, Joseph A., 1979– author. |
Winter, David A., 1930-2012, author. | Winter, David A., 1930-2012.
Biomechanics and motor control of human movement.
Title: Winter’s biomechanics and motor control of human movement / Stephen
J. Thomas (Thomas Jefferson University), Joseph A. Zeni (Rutgers
University), David A. Winter (Waterloo, Ontario, Canada).
Other titles: Biomechanics and motor control of human movement
Description: Fifth edition. | Hoboken, NJ : Wiley, 2023. | Includes
bibliographical references and index.
Identifiers: LCCN 2022019974 (print) | LCCN 2022019975 (ebook) | ISBN
9781119827023 (cloth) | ISBN 9781119827030 (pdf) | ISBN 9781119827047
(epub)
Subjects: LCSH: Human mechanics. | Motor ability. | Kinesiology.
Classification: LCC QP303 .T46 2023 (print) | LCC QP303 (ebook) | DDC
612.7/6–dc23/eng/20220720
LC record available at https://lccn.loc.gov/2022019974
LC ebook record available at https://lccn.loc.gov/2022019975
Cover Image: © John Emerson

Cover Design: Wiley
Set in 10.75/12pt Times by Straive, Pondicherry, India.
ISTUDY
DEDICATION
We would like to dedicate this edition to David A. Winter and his family. To those inside the field
of biomechanics, David Winter needs no introduction. He is often described as the founding
father of modern biomechanics and rightfully so. His education was in the field of electrical
engineering and he went on to become a Fellow of the Institute of Electrical and Electronics
Engineers. However, his education, training, and service were not constrained to electrical
engineering. He served in professorship roles in biomedical engineering, surgery, and kinesiol-
ogy. He was a founding member of the Canadian Society for Biomechanics and later, became
the first recipient of the Career Investigators Award by the same society. He received the
Lifetime Achievement Award by the Gait and Clinical Movement Analysis Society, as well
as the Muybridge Medal by the International Society of Biomechanics, which is the society’s
most prestigious honor. He solely authored four previous editions of this textbook, which is con-
sidered the seminal technical biomechanics textbook and is still readily used by professionals in
biomechanics, engineering, kinesiology, rehabilitation sciences, and many other related fields.
He also authored three additional textbooks and over 100 peer-reviewed research publications.
He finished his career as a Distinguished Professor Emeritus in Kinesiology at the University of
Waterloo.
His accomplishments in the field of biomechanics demonstrate his drive and passion for his
profession. However, his passion and love for his family never took a backseat to his career. We
were very fortunate to meet and speak with his three children Merriam, Andrew, and Bruce. The
stories they shared about their father made clear that he was a constant and loving presence in
their lives. The Winter family always ate dinner together, which served as a roundtable for
insightful and thought-provoking conversations about the things the children were currently
involved in at school and life. His children revealed that their father never missed an event or
activity in their busy lives. His daughter Merriam shared a memory of her father from when
she was six years old. When she happened upon her father working in the basement, she asked
“What are you doing, dad?” He responded “Making wood!” What he really was doing was build-
ing a life-size playhouse for the backyard that included a porch, flower boxes in the windows,
benches, a table, and all painted green and pink. She also shared her memories about how David
Winter built the family a cottage on the outskirts of Halifax from scratch. His son Bruce shared
stories of visiting his father in his gait lab when he was attending college and being able to see
David Winter’s passion for teaching students and working with patients. David Winter poured his
devotion to teaching his students, and Bruce shared that these students became part of the family
ISTUDY
and would often join the family dinner at their house. These heartwarming stories demonstrate
that it is possible to achieve so much in your career, while still putting your family in the forefront.
This is an important lesson to anyone in academia or industry, vocations which continually com-
pete for our precious time.
In speaking with David Winter’s family, a key lesson became clear; the foundation for a
successful life and career is to do what you love, while supporting those around you; whether
that be your patients, students, colleagues, or most importantly, your family and friends. We
would like to thank his children and grandchildren for sharing their dad and grandfather with
all of us, in the field of biomechanics, as we are all better because of him.
STEPHEN J. THOMAS, PHD, ATC

JOSEPH A. ZENI, PHD, PT
ISTUDY
CONTENTS
List of Contributors xv
Preface xvii
Acknowledgments xix
About the Companion Website xxi
1 Biomechanics as an Interdiscipline 1
Stephen J. Thomas, Joseph A. Zeni, and David A. Winters
1.0 Introduction, 1
1.0.1 Importance of Human Movement Analysis, 1
1.0.2 The Interprofessional Team, 2
1.1 Measurement, Description, Analysis, and Assessment, 2
1.1.1 Measurement, Description, and Monitoring, 3
1.1.2 Analysis, 4
1.1.3 Assessment and Interpretation, 5
1.2 Biomechanics and its Relationship with Physiology and Anatomy, 6
1.3 References, 7
2 Signal Processing 8
Joseph A. Zeni, Stephen J. Thomas, and David A. Winters
2.0 Introduction, 8
2.1 Auto- and Cross-Correlation Analyses, 8
2.1.1 Similarity to the Pearson Correlation, 9
2.1.2 Formulae for Auto- and Cross-Correlation Coefficients, 10
2.1.3 Four Properties of the Autocorrelation Function, 11
2.1.4 Three Properties of the Cross-Correlation Function, 14
2.1.5 Importance in Removing the Mean Bias from the Signal, 15
2.1.6 Digital Implementation of Auto- and Cross-Correlation Functions, 15
ISTUDY
viii CONTENTS
2.1.7 Application of Autocorrelations, 16

2.1.8 Applications of Cross-Correlations, 17
2.2 Frequency Analysis, 19
2.2.1 Introduction – Time Domain vs. Frequency Domain, 19
2.2.2 Discrete Fourier (Harmonic) Analysis, 19
2.2.3 Fast Fourier Transform (FFT), 21
2.2.4 Applications of Spectrum Analyses, 22
2.3 Ensemble Averaging of Repetitive Waveforms, 29
2.3.1 Examples of Ensemble-Averaged Profiles, 31
2.3.2 Normalization of Time Bases to 100%, 31
2.3.3 Measure of Average Variability about the Mean Waveform, 32
2.4 References, 32
3 Kinematics 34
Amy L. Lenz
3.0 Historical Development and Complexity of Problem, 34

3.1 Kinematic Conventions, 35
3.1.1 Absolute Spatial Reference System, 35
3.1.2 Total Description of a Body Segment in Space, 36
3.2 Direct Measurement Techniques, 36
3.2.1 Goniometers, 36
3.2.2 Accelerometers, 38
3.2.3 Inertial Sensors, 39
3.2.4 Special Joint Angle Measuring Systems, 40
3.2.5 Electromagnetic Systems, 41
3.3 Imaging Measurement Techniques, 42
3.3.1 Review of Basic Lens Optics, 42
3.3.2 f-Stop Setting and Field of Focus, 43
3.3.3 Television Imaging Camera Historical Development, 43
3.3.4 Optical Motion Capture, 44
3.3.5 Optoelectric Techniques, 47
3.3.6 Biplane Fluoroscopy, 48
3.3.7 Markerless Systems, 51
3.3.8 Summary of Various Kinematic Systems, 51
3.4 Clinical Measures of Kinematics, 52
3.4.1 2-D Kinematic Apps/Sensors, 52
3.4.2 Sensor-Based Systems, 52
3.5 Processing of Raw Kinematic Data, 52
3.5.1 Nature of Unprocessed Image Data, 52
3.5.2 Signal Versus Noise in Kinematic Data, 53
3.5.3 Problems of Calculating Velocities and Accelerations, 54
3.5.4 Smoothing and Curve Fitting of Data, 54
3.5.5 Comparison of Some Smoothing Techniques, 60
3.6 Calculation of Other Kinematic Variables, 62
3.6.1 Limb-Segment Angles, 62
3.6.2 Joint Angles, 63
ISTUDY
CONTENTS ix
3.6.3 Velocities – Linear and Angular, 63

3.6.4 Accelerations – Linear and Angular, 63
3.7 Problems Based on Kinematic Data, 64
3.8 References, 65
4 Anthropometry 67
4.0 Scope of Anthropometry in Movement Biomechanics, 67
4.0.1 Segment Dimensions, 67
4.1 Density, Mass, and Inertial Properties, 68
4.1.1 Whole-Body Density, 68
4.1.2 Segment Densities, 69
4.1.3 Segment Mass and Center of Mass, 69
4.1.4 Center of Mass of a Multisegment System, 72
4.1.5 Mass Moment of Inertia and Radius of Gyration, 73
4.1.6 Parallel Axis Theorem, 74
4.1.7 Use of Anthropometric Tables and Kinematic Data, 75
4.2 Direct Experimental Measures, 78
4.2.1 Location of the Anatomical Center of Mass of the Body, 79
4.2.2 Calculation of the Mass of a Distal Segment, 79
4.2.3 Moment of Inertia of a Distal Segment, 80
4.2.4 Joint Axes of Rotation, 81
4.3 Muscle Anthropometry, 82
4.3.1 Cross-Sectional Area of Muscles, 82
4.3.2 Change in Muscle Length During Movement, 83
4.3.3 Force per Unit Cross-Sectional Area (Stress), 84
4.3.4 Mechanical Advantage of Muscle, 84
4.3.5 Multijoint Muscles, 85
4.4 Problems Based on Anthropometric Data, 86
4.5 References, 87
5 Kinetics: Forces and Moments of Force 89

5.0 Biomechanical Models, 89
5.0.1 Link-Segment Model Development, 89
5.0.2 Forces Acting on the Link-Segment Model, 90
5.0.3 Joint Reaction Forces and Bone-on-Bone Forces, 91
5.1 Basic Link-Segment Equations – The Free-Body Diagram, 93
5.2 Force Transducers and Force Plates, 98
5.2.1 Multidirectional Force Transducers, 98
5.2.2 Force Plates, 99
5.2.3 Combined Force Plate and Kinematic Data, 104
5.2.4 Interpretation of Moment-of-Force Curves, 105
5.2.5 Differences Between Center of Mass and Center of Pressure, 107
5.2.6 Kinematics and Kinetics of the Inverted Pendulum Model, 108
ISTUDY
x CONTENTS
5.3 Bone-on-bone Forces During Dynamic Conditions, 110

5.3.1 Indeterminacy in Muscle Force Estimates, 110
5.3.2 Example Problem, 111
5.4 References, 114
6 Mechanical Work, Energy, and Power 115
6.0 Introduction, 115

6.0.1 Mechanical Energy and Work, 115
6.0.2 Law of Conservation of Energy, 116
6.0.3 Internal Versus External Work, 116
6.0.4 Positive Work of Muscles, 118
6.0.5 Negative Work of Muscles, 118
6.0.6 Muscle Mechanical Power, 119
6.0.7 Mechanical Work of Muscles, 119
6.0.8 Mechanical Work Done on an External Load, 120
6.0.9 Mechanical Energy Transfer Between Segments, 122
6.1 Efficiency, 123
6.1.1 Causes of Inefficient Movement, 124
6.1.2 Summary of Energy Flows, 127
6.2 Forms of Energy Storage, 128
6.2.1 Energy of a Body Segment and Exchanges of Energy Within the
Segment, 129
6.2.2 Total Energy of a Multisegment System, 132
6.3 Calculation of Internal and External Work, 133
6.3.1 Internal Work Calculation, 133
6.3.2 External Work Calculation, 136
6.4 Power Balances at Joints and Within Segments, 136
6.4.1 Energy Transfer via Muscles, 137
6.4.2 Power Balance Within Segments, 138
6.5 Problems Based on Kinetic and Kinematic Data, 141
6.6 References, 143
7 Understanding 3D Kinematic and Kinetic Variables 145

Thomas Hulcher

7.1 Axes Systems, 145
7.1.1 Global Reference System, 145
7.1.2 Local Reference Systems and Rotation of Axes, 146
7.1.3 Other Possible Rotation Sequences, 147
7.1.4 Dot and Cross Products, 148
7.2 Marker and Anatomical Axes Systems, 148
7.2.1 Markerset Design, 150
7.2.2 Event Detection Methods for Gait, 152
7.2.3 Event Detection Methods for Other Activities, 153
7.2.4 Considerations for Applications with
7.2.5 Example of a Kinematic Data Set, 154
ISTUDY
CONTENTS xi
7.3 Determination of Segment Angular Velocities and Accelerations, 158

7.4 Kinetic Analysis of Reaction Forces and Moments, 162
7.4.1 Newtonian Three-Dimensional Equations of Motion for a Segment, 162
7.4.2 Euler’s Three-Dimensional Equations of Motion for a Segment, 163
7.4.3 Example of a Kinetic Data Set, 164
7.4.4 Joint Mechanical Powers, 167
7.4.5 Induced Acceleration Analysis, 167
7.4.6 Sample Moment and Power Curves, 168
7.5 Suggested Further Reading, 170
7.6 References, 170
8 Muscle Mechanics 171

8.0.1 The Motor Unit, 171
8.0.2 Recruitment of Motor Units, 172
8.0.3 Size Principle, 173
8.0.4 Types of Motor Units – Fast- and Slow-Twitch Classification, 174
8.0.5 The Muscle Twitch, 175
8.0.6 Shape of Graded Contractions, 176
8.1 Force–Length Characteristics of Muscles, 177
8.1.1 Force–Length Curve of the Contractile Element, 177
8.1.2 Influence of Parallel Connective Tissue, 178
8.1.3 Series Elastic Tissue, 178
8.1.4 In Vivo Force–Length Measures, 180
8.2 Force–Velocity Characteristics, 181
8.2.1 Concentric Contractions, 181
8.2.2 Eccentric Contractions, 183
8.2.3 Combination of Length and Velocity Versus Force, 183
8.2.4 Combining Muscle Characteristics with Load Characteristics:
Equilibrium, 184
8.3 Technique to Measure in Vivo Tendon Mechanical Properties, 186
8.3.1 Ankle Joint Moment, 186
8.3.2 Tendon Mechanical Properties, 187
8.4 References, 187
9 Kinesiological Electromyography 189


9.1 Electrophysiology of Muscle Contraction, 189
9.1.1 Motor End Plate, 189
9.1.2 Sequence of Chemical Events Leading to a Twitch, 190
9.1.3 Generation of a Muscle Action Potential, 190
9.1.4 Duration of the Motor Unit Action Potential, 192
9.1.5 Detection of Motor Unit Action Potentials from Electromyogram During
Graded Contractions, 194
ISTUDY
xii CONTENTS
9.2 Recording of the Electromyogram, 195

9.2.1 Amplifier Gain, 196
9.2.2 Input Impedance, 196
9.2.3 Frequency Response, 197
9.2.4 Common-Mode Rejection, 199
9.2.5 Cross-Talk in Surface Electromyograms, 202
9.2.6 Recommendations for Surface Electromyogram Reporting and Electrode
Placement Procedures, 205
9.3 Processing of the Electromyogram, 205
9.3.1 Full-Wave Rectification, 206
9.3.2 Linear Envelope, 207
9.3.3 True Mathematical Integrators, 208
9.4 Relationship Between Electromyogram and Biomechanical Variables, 208
9.4.1 Electromyogram Versus Isometric Tension, 209
9.4.2 Electromyogram During Muscle Shortening and Lengthening, 210
9.4.3 Electromyogram Changes During Fatigue, 211
9.5 References, 212
10 Modeling of Human Movement 215

Brian A. Knarr, Todd J. Leutzinger, and Namwoong Kim

10.1 Review of Forward Solution Models, 216
10.1.1 Assumptions and Constraints of Forward Solution Models, 217
10.1.2 Potential of Forward Solution Simulations, 217
10.2 Muscle-Actuated Simulation of Movement, 218
10.2.1 Musculoskeletal Modeling, 218
10.2.2 Control, 221
10.2.3 OpenSim, 223
10.2.4 EMG-Driven Modeling, 227
10.3 Model Validation, 230
10.4 References, 231
11 Static and Dynamic Balance 235


11.1 The Support Moment Synergy, 236
11.1.1 Relationship Between Ms and the Vertical Ground Reaction Force, 237
11.2 Medial/Lateral and Anterior/Posterior Balance in Standing, 239
11.2.1 Quiet Standing, 239
11.2.2 Medial Lateral Balance Control During Workplace Tasks, 240
11.3 Dynamic Balance During Walking, 241
11.3.1 The Human Inverted Pendulum in Steady State Walking, 241
11.3.2 Initiation of Gait, 242
11.3.3 Gait Termination, 244
ISTUDY
CONTENTS xiii
12 Central Nervous System’s Role in Biomechanics 247

Alan R. Needle and Christopher J. Burcal
12.1 Central Nervous System and Volitional Control of Movement, 247
12.1.1 Key Structures for Movement, 247
12.1.2 Synapses and Neurotransmitters, 249
12.1.3 CNS Adaptations, 249
12.2 Peripheral Nervous System and Reflexive Control of Movement, 250
12.2.1 Sensory Receptors and Motor Units, 252
12.3 Methodologies to Understand Central Nervous System Function, 253
12.3.1 Functional Magnetic Resonance Imaging (fMRI), 253
12.3.2 Electroencephalography (EEG), 257
12.3.3 Neural Excitability, 265
12.4 Peripheral Nervous System Measurement Techniques, 269
12.4.1 Nerve Conduction Studies, 269
12.4.2 Microneurography, 271
12.5 Methodologies to Understand Central Nervous System Behavior and Environmental
Interactions, 271
12.5.1 Virtual Reality, 271
12.6 Nervous System Role in Muscle Synergies, 274
12.6.1 Measurement Techniques and Experimental Setup, 274
12.6.2 Analysis Techniques, 275
12.7 The Central Nervous System and Learning, and Injury, 276
12.7.1 Translation of Synaptic Plasticity to Motor Learning, 276
12.7.2 Role of Pathology on the Central Nervous System, 276
13 A Case-Based Approach to Interpreting Biomechanical Data 281

Ankur Padhye, John D. Willson, Joseph A. Zeni, Kristen F. Nicholson, and Garrett S. Bullock
13.0 Patellofemoral Pain, 281
13.0.1 Introduction, 281
13.0.2 Case Description, 281
13.0.3 Patient Examination, 282
13.0.4 Gait Analysis, 282
13.0.5 Interpretations and Intervention, 282
13.0.6 Patient Outcomes and Discussion, 283
13.0.7 Conclusion, 284
13.0.8 References, 284
13.1 Biomechanical Approach to Manage Knee Osteoarthritis, 284
13.1.1 Osteoarthritis and Biomechanics, 284
13.1.2 Patient History, 286
13.1.3 Biomechanical Assessment, 286
13.2 Ulnar Collateral Ligament Reconstruction, 288
13.2.1 Player History, 289
ISTUDY
xiv CONTENTS
APPENDICES
A. Kinematic, Kinetic, and Energy Data 295

Figure A.1 Walking Trial – Marker Locations and Mass and Frame Rate
Information, 295
Table A.1 Raw Coordinate Data (cm), 296
Table A.2(a) Filtered Marker Kinematics – Rib Cage and Greater Trochanter (Hip), 300
Table A.2(b) Filtered Marker Kinematics – Femoral Lateral Epicondyle (Knee) and
Head of Fibula, 304
Table A.2(c) Filtered Marker Kinematics – Lateral Malleolus (Ankle) and Heel, 308
Table A.2(d) Filtered Marker Kinematics – Fifth Metatarsal and Toe, 312
Table A.3(a) Linear and Angular Kinematics – Foot, 316
Table A.3(b) Linear and Angular Kinematics – Leg, 320
Table A.3(c) Linear and Angular Kinematics – Thigh, 324
Table A.3(d) Linear and Angular Kinematics – ½ HAT, 328
Table A.4 Relative Joint Angular Kinematics – Ankle, Knee, and Hip, 332
Table A.5(a) Reaction Forces and Moments of Force – Ankle and Knee, 336
Table A.5(b) Reaction Forces and Moments of Force – Hip, 340
Table A.6 Segment Potential, Kinetic, and Total Energies – Foot, Leg, Thigh, and
½ HAT, 344
Table A.7 Power Generation/Absorption and Transfer – Ankle, Knee, and Hip, 348
B. Units and Definitions Related to Biomechanical and Electromyographical

Measurements 351
Table B.1 Base SI Units, 351
Table B.2 Derived SI Units, 352
Index 355
ISTUDY
LIST OF CONTRIBUTORS
Garrett S. Bullock, PT, DPT, DPhil, Instructor, Department of Orthopaedic Surgery, Wake
Forest School of Medicine, Winston-Salem, NC, USA
Christopher J. Burcal, PhD, ATC, Assistant Professor, School of Health and Kinesiology, Uni-
versity of Nebraska at Omaha, Omaha, NE, USA
Thomas Hulcher, M.S., Biomechanist, Department of Physical Therapy, Jefferson College of

Rehabilitation Science, Thomas Jefferson University, Philadelphia, PA, USA
Namwoong Kim, PhD, Research Associate, Division of Health & Kinesiology, Incheon
National University, Incheon, Republic of Korea
Brian A. Knarr, PhD, Associate Professor, Department of Biomechanics, College of Education,

Health, and Human Sciences, University of Nebraska at Omaha, Omaha, NE, USA
Amy L. Lenz, PhD, Research Instructor, Department of Orthopaedics, School of Medicine, Uni-
versity of Utah, Salt Lake City, UT, USA
Todd J. Leutzinger, PhD, Instructor, Department of Biomechanics, College of Education,

Health, and Human Sciences, University of Nebraska at Omaha, Omaha, NE, USA
Alan R. Needle, PhD, ATC, LAT, CSCS, Associate Professor, Department of Public Health &
Exercise Science and Department of Rehabilitation Science, Appalachian State University,
Boone, NC, USA.
Kristen F. Nicholson, PhD, Assistant Professor, Department of Orthopaedic Surgery, Wake

Forest School of Medicine, Winston-Salem, NC, USA
Ankur Padhye PT, PhD Student, Rehabilitation Sciences/Movement Science and Disorders
Concentration, Department of Physical Therapy, East Carolina University, Greenville, NC, USA
ISTUDY
xvi LIST OF CONTRIBUTORS
Stephen J. Thomas, PhD, ATC, Associate Professor and Department Chair, Department of
Exercise Science, Jefferson College of Rehabilitation Science, Thomas Jefferson University,
Philadelphia, PA, USA
John D. Willson, PhD, PT, Associate Professor, Department of Physical Therapy, East Carolina
University, Greenville, NC, USA
David A. Winters (Deceased), University of Waterloo, Waterloo, Ontario, Canada
Joseph A. Zeni, PhD, PT, Associate Professor, Department of Rehabilitation and Movement
Sciences, School of Health Professions, Rutgers University, Newark, NJ, USA
ISTUDY
PREFACE
The field of biomechanics has experienced a wealth of growth since the printing of the fourth
edition. With sweeping technological advancements, biomechanics has become an integral com-
ponent of robotics, healthcare and medicine, athletic performance, movies and animation, and
numerous other fields. With these changes, the discipline of biomechanics has become more tan-
gible, recognizable, and accessible to individuals across the world. This new fifth edition reflects
the expanded breadth and depth of biomechanics and includes updated references and research.
The original text, Biomechanics of Human Movement, published in 1979, had its title changed
to Biomechanics and Motor Control of Human Movement when the second edition was published
in 1990. This change is to acknowledge the new directions and research of the 1980s. In that
second edition, five of eight chapters addressed various aspects of muscles and motor systems.
The third edition, published in 2004, added information about three-dimensional (3D) kinematics
and kinetics, reflecting the continued emphasis on motor control and the close links between bio-
mechanics and movement analysis. The fourth edition, published in 2009, added updated discus-
sions of anatomy, muscle physiology, and electromyography, with a strong emphasis on dynamic
movements and in vivo data and analyses.
In keeping with the past updates, this updated edition integrates the new directions of biome-
chanics. These new directions include a strong emphasis on the role of the central nervous system
in motor control and biomechanical analyses, as well as a focus on practical applications of bio-
mechanics. The fifth edition has two new chapters, Chapter 12, “Central Nervous System’s Role in
Biomechanics,” and Chapter 13, “A Case-Based Approach to Interpreting Biomechanical Data.”
Chapter 3, “Kinematics,” has undergone significant updates to incorporate new technology and
approaches to measuring human movement, including, but not limited to markerless technology.
Chapter 7, “Understanding 3-D Kinematic and Kinetic Variables,” includes new sections on mar-
kersets design, event detection, the incorporation of ISB standards, and induced acceleration anal-
ysis. Chapter 10, “Modeling of Human Movement” has been completely rewritten to include the
current modeling approaches and technologies used in biomechanical research. Chapter 11 has
been renamed to “Static and Dynamic Balance” to better reflect the content in this chapter. All chap-
ters have also undergone edits to include updated information in each of the topic areas. The appen-
dices, which underwent major additions in the second edition, remain intact. The extensive
numerical tables contained in Appendix A: “Kinematic, Kinetic, and Energy Data” can also be
found online in excel format to facilitate learning experiences in the modern classroom.
ISTUDY
xviii PREFACE
As was stated in the original editions, it is expected that the student has had basic courses in
anatomy, mechanics, calculus, and electrical science. The major disciplines to which the book is
directed are kinesiology, biomechanics, engineering (bioengineering, mechanical, and rehabili-
tation engineering), physical education, ergonomics, physical, and occupational therapy, athletic
training, neuroscientists, sport scientists, and exercise scientists. The text should also prove val-
uable to researchers in orthopedics, muscle physiology, and rehabilitation medicine. Lastly, this
text can provide valuable information to athletic coaches to help guide them in the appropriate use
of biomechanical data to make training and coaching decisions.
S TEPHEN J. T HOMAS , P H D, ATC

Thomas Jefferson University
March 2022
J OSEPH A. Z ENI , P H D, PT
Rutgers University
Newark, NJ, USA
March 2022
ISTUDY
ACKNOWLEDGMENTS
My first acknowledgment goes to my family. To my parents, Howard and Carol for instilling in
me the Scranton, PA grit, and work ethic that has guided me through my career. To my wife
Regina, the love of my life, for always supporting our family, and making us laugh. To my
son Jaxson and daughter Penelope for always making me proud and putting a smile on my face.
My second acknowledgment goes to my mentors, colleagues, and students who have guided
me, inspired me, and helped me tremendously along the way. Although I cannot include everyone
I would like to acknowledge a few. Buz and Kathy Swanik who were my mentors from undergrad
through my PhD and played a significant role in my career. John Kelly has been a lifelong mentor
and big brother not only in my career by also my personal life. Katie Reuther has been a great
friend and colleague who taught me many engineering principles starting during my postdoc at
the University of Pennsylvania. Tommy Nowakowski, one of my best friends, who is responsible
for me becoming an Athletic Trainer. My first cohort of students (Justin Cobb, Ohsana Valle,
Morgan Wambold, and Ryan Paul) who I am proud to say I have learned more from then they
learned from me. Lastly, all of the professors in the Biomechanics PhD program at the University
of Delaware.
Without all of you, this would not be possible!
S TEPHEN J. T HOMAS , P H D, ATC
This work would not be possible without those who have supported, educated, and nurtured me
throughout my life and career. I would like to acknowledge the consistent support of my parents,
Kathleen and Joe, who have always worked tirelessly to provide me with opportunities to
succeed. My wife, Courtney, has been an unwavering rock in our household, balancing our fam-
ily, her work, and my endeavors for the past 15 years. I must also acknowledge those who have
not only taught me foundational biomechanics, but also passed along a passion for research, edu-
cation, and the academic world. Foremost, my PhD advisor, Jill Higginson, a biomechanical
engineer who took a chance and welcomed me, a clinician with little to no understanding of
mechanical systems, into her lab. I learned so much during that time and fell in love with the
application of engineering principles to the human body. To Anil Bhave, who introduced me
ISTUDY
xx ACKNOWLEDGMENTS
to the clinical world of motion analysis, I am forever grateful for your enthusiasm and patience as
my mentor and boss. Numerous professors from the University of Delaware Biomechanics and
Movement Science program played a critical role and pushed me to be a better student and enjoy
the process of learning: Jim Richards, Bill Rose, Lynn Snyder-Mackler, Thomas Buchanan,
Stuart Binder-Macleod, and many others; I thank you for everything. These professors are
educators, mentors, and colleagues without equal.
J OSEPH A. Z ENI , P H D, PT
ISTUDY
ABOUT THE COMPANION WEBSITE
This book is accompanied by a companion website
www.wiley.com/go/Thomas/WintersBiomechanics5e
The website includes:

• Powerpoints which contains the key point for all the individual chapters.
• Appendix contains table regarding the kinematic, kinetic, and energy.
ISTUDY
ISTUDY
1
BIOMECHANICS AS AN INTERDISCIPLINE
STEPHEN J. THOMAS1, JOSEPH A. ZENI2, AND DAVID A. WINTERS3,†
1
2
3
1.0 INTRODUCTION
Biomechanics of human movement is a field that has grown and advanced significantly in the
past decade and includes observing, measuring, analyzing, assessing, and interpreting human
movement. A wide variety of physical movements are involved – everything from the gait of
the physically handicapped to the lifting of a load by a factory worker to the performance of
a superior athlete. The physical and biological principles that apply are the same in all cases. What
changes from case to case are the specific movement tasks and the level of detail that is being
asked about the assessment and interpretation of each movement.
The list of professionals interested in biomechanics is quite long: orthopedic surgeons, athletic
trainers, biomedical and mechanical engineers, occupational and physical therapists, kinesiolo-
gists, sport scientists, prosthetists, psychiatrists, orthotists, athletic coaches, sports equipment
designers, and so on. At the basic level, the name given to the science dedicated to the area
of human movement is kinesiology. It is a broad discipline blending aspects of psychology, sports
nutrition, motor development and learning, and exercise physiology as well as biomechanics.
Biomechanics, as an outgrowth of both life and physical sciences, is built on the basic body
of knowledge of physics, chemistry, mathematics, physiology, and anatomy. It is amazing to
note that the first real “biomechanicians” date back to Leonardo da Vinci, Galileo, Lagrange,
Bernoulli, Euler, and Young. All these scientists had primary interests in the application of
mechanics to biological problems.
1.0.1 Importance of Human Movement Analysis

The first question that one may ask is “What is the benefit of assessing and interpreting human
movement.” The answer to this can vary depending on the specific movement being studied and
the expected outcomes for that specific individual. At the most basic level, it is important to
†
Deceased.
Winter’s Biomechanics and Motor Control of Human Movement, Fifth Edition.

Stephen J. Thomas, Joseph A. Zeni, and David A. Winter.
© 2023 John Wiley & Sons, Inc. Published 2023 by John Wiley & Sons, Inc.
Companion website:
ISTUDY
2 BIOMECHANICS AS AN INTERDISCIPLINE
understand the underlying mechanisms responsible for the development of movement and com-
pensation due to injury or pain. These mechanisms will act as a roadmap or equation that clin-
icians can utilize during rehabilitation to optimize movement, which will result in long-term
recovery and injury prevention. When asking this question in reference to athletics, the goal
is typically to increase performance and also to prevent overuse injuries. The current conundrum
in many sports is that as performance increases the risk of injury also increases. This can result in
an exponentially challenging situation for biomechanists to resolve when working with athletes.
1.0.2 The Interprofessional Team

As the field of biomechanics continues to evolve it is becoming much more evident that a team
approach needs to be used when working with patients/clients. As mentioned previously, the pro-
fessions interested in human movement are very broad and span many disciplines that have
unique skill sets to help patients/clients. This includes surgery, rehabilitation, exercise, mental
health, nutrition, equipment and prosthetic designs, etc. It is unfortunate that often times these
professions work in silos, which often result in conflicting advice and direction. This not only
confuses the patients/clients but also causes set backs in their progression. The first step in devel-
oping an interprofessional team is identifying and respecting the similarities and differences in
expertise across the continuum of care for the patient/client. The focus always needs to be direc-
ted toward the patient/client. The next step is effective communication between team members
and the patient/client. Each team member needs to communicate collectively following assess-
ments so that all of the collected information can be discussed and interpreted. This interpretation
will then be used by the team to create an optimal plan to achieve the patient’s/client’s goals. It is
also very important to communicate this plan to the patient/client in a digestible form. Educa-
tional programs in biomechanics often focus on the science and technology used to measure
and assess human movement very accurately. However, it often falls short on providing training
for interpreting and communicating the results of human movement assessments to the patient/
client. This is an incredibly important and valuable skill that needs to be taught in educational
programs, which will result in the expansion of the field outside the traditional settings of research
labs and provide value to patients/clients from all walks of life.
1.1 MEASUREMENT, DESCRIPTION, ANALYSIS, AND ASSESSMENT
The scientific approach as applied to biomechanics has been characterized by a fair amount of con-
fusion. Some descriptions of human movement have been passed off as assessments, some studies
involving only measurements have been falsely advertised as analyses, and so on. It is, therefore,
important to clarify these terms. Any quantitative assessment of human movement must be pre-
ceded by a measurement and description phase, and if more meaningful diagnostics are needed, a
biomechanical analysis is usually necessary. Most of the material in this text is aimed at the tech-
nology of measurement and description and the modeling process required for analysis. The final
interpretation, assessment, or diagnosis is movement specific and is limited to the examples given.
Figure 1.1, which has been prepared for the assessment of the physically handicapped, depicts
the relationships between these various phases of assessment. All levels of assessment involve a
human being and are based on his or her visual observation of a patient or subject, recorded data,
or some resulting biomechanical analysis. The primary assessment level uses direct observation,
which places tremendous “overload” even on the most experienced observer. All measures are
subjective and are almost impossible to compare with those obtained previously. Observers are
then faced with the tasks of documenting (describing) what they see, monitoring changes, ana-
lyzing the information, and diagnosing the causes. If measurements can be made during the
patient’s movement, then data can be presented in a convenient manner to describe the movement
quantitatively. Here the assessor’s task is considerably
ISTUDY
1.1 MEASUREMENT, DESCRIPTION, ANALYSIS, AND ASSESSMENT 3
Assessment
Data
Describe
monitor
analyze
diagnose
Previous
patient
data Data
on
Data normals
Measure Describe
Monitor
analyze
diagnose
Analyze
Diagnose
Previous Analysis
patient on
analysis normals
Figure 1.1 Schematic diagram showing the three levels of assessment of human movement.
changes, carry out simple analyses, and try to reach a more objective diagnosis. At the highest
level of assessment, the observer can view biomechanical analyses that are extremely powerful in
diagnosing the exact cause of the problem, compare these analyses with the normal population,
and monitor their detailed changes with time.
The measurement and analysis techniques used in an athletic event could be identical to the
techniques used to evaluate an amputee’s gait. However, the assessment of the optimization of the
energetics of the athlete is quite different from the assessment of the stability of the amputee.
Athletes are looking for very detailed but minor changes that will improve their performance
by a few percentage points, sufficient to move them from fourth to first place. Their training
and exercise programs and reassessment normally continue over an extended period of time.
The amputee, on the other hand, is looking for major improvements, probably related to safe
walking, but not fine and detailed differences. This person is quite happy to be able to walk
at less than maximum capability, although techniques are available to permit training and have
the prosthesis readjusted until the amputee reaches some perceived maximum. When working
with Para-athletes these approaches and goals are often are blended together. In ergonomic stud-
ies, assessors are likely looking for maximum stresses in specific tissues during a given task, to
thereby ascertain whether the tissue is working within safe limits. If not, they will analyze pos-
sible changes in the workplace or task in order to reduce the stress or fatigue.
1.1.1 Measurement, Description, and Monitoring

It is difficult to separate the two functions of measurement and description. However, for clarity,
the student should be aware that a given measurement device can have its data presented in a
number of different ways. Conversely, a given description could have come from several differ-
ent measurement devices.
Earlier biomechanical studies had the sole purpose of describing a given movement, and any
assessments that were made resulted from visual inspection of the data. The description of the
data can take many forms: plots of body coordinates, stick
ISTUDY
such as gait velocity, load lifted, or height of a jump. A video camera, by itself, is a measurement
device, and the resulting plots form the description of the event in time and space. The coordi-
nates of key anatomical landmarks can be extracted and plotted at regular intervals in time. Time
history plots of one or more coordinates are useful in describing detailed changes in a particular
landmark. They also can reveal changes in velocity and acceleration. A total description in the
plane of the movement is provided by the stick diagram, in which each body segment is repre-
sented by a straight line or stick. Joining the sticks together gives the spatial orientation of all
segments at any point in time. Repetition of this plot at equal intervals of time gives a pictorial
and anatomical description of the dynamics of the movement. Here, trajectories, velocities, and
accelerations can be visualized. To get some idea of the volume of the data present in a stick
diagram, the student should note that one full page of coordinate data is required to make the
complete plot for the description of the event. The coordinate data can be used directly for
any desired analysis: reaction forces, muscle moments, energy changes, efficiency, and so on.
Conversely, an assessment can occasionally be made directly from the description. A trained
observer, for example, can scan a stick diagram and extract useful information that will give some
directions for training or therapy, or give the researcher some insight into basic mechanisms of
movement.
The term monitor needs to be introduced in conjunction with the term describe. To monitor
means to note changes over time. Thus, a physical therapist will monitor the progress (or the lack
of it) for each physically disabled person undergoing therapy. Only through accurate and reliable
measurements will the therapist be able to monitor any improvement and thereby make inferences
to the validity of the current therapy. What monitoring does not tell us is why an improvement is
or is not taking place; it merely documents the change. All too many coaches or therapists doc-
ument the changes with the inferred assumption that their intervention has been the cause. How-
ever, the scientific rationale behind such inferences is missing. Unless a detailed analysis is done,
we cannot document the detailed motor-level changes that will reflect the results of therapy or
training.
1.1.2 Analysis
The measurement system yields data that are suitable for analysis. This means that data have been
calibrated and are as free as possible from noise and artifacts. Analysis can be defined as any math-
ematical operation that is performed on a set of data to present them in another form or to combine
the data from several sources to produce a variable that is not directly measurable. From the ana-
lyzed data, information may be extracted to assist in the assessment stage. In some cases, the math-
ematical operation can be very simple, such as the processing of an electromyographic (EMG)
signal to yield an envelope signal. The mathematical operation performed here can be described
in two stages. The first is a full-wave rectifier (the electronic term for a circuit that gives the abso-
lute value). The second stage is a low-pass filter (which mathematically has the same transfer func-
tion as that between a neural pulse and its resultant muscle twitch). A more complex biomechanical
analysis could involve a link-segment model, and with appropriate kinematic, anthropometric, and
kinetic output data, we can carry out analyses that could yield a multitude of significant time-course
curves. Figure 1.2 depicts the relationships between some of these variables. The output of the
movement is what we see. It can be described by a large number of kinematic variables: displace-
ments, joint angles, velocities, and accelerations. If we have an accurate model of the human body
in terms of anthropometric variables, we can develop a reliable link-segment model. With this
model and accurate kinematic data, we can predict the net forces and muscle moments that caused
the movement we just observed. Such an analysis technique is called an inverse solution. It is
extremely valuable, as it allows us to estimate variables such as joint reaction forces and moments.
Such variables are not measurable directly. In a similar manner, individual muscle forces might be
predicted through the development of a mathematical model of a muscle, which could have neural
drive, length, velocity, and cross-sectional area as inputs.
ISTUDY
1.1 MEASUREMENT, DESCRIPTION, ANALYSIS, AND ASSESSMENT 5
M. Forces ë
EMG ė
M. Moments Link e
Muscle Ẋ˙, Ẏ˙
Segment
Neural Model External Model Ẋ, Ẏ
Stimul. forces X, Y
Mech. energy
Figure 1.2 Schematic diagram to show the relationship between the neural, kinetic, and kinematic variables
required to describe and analyze human movement.
1.1.3 Assessment and Interpretation

The entire purpose of any assessment is to make a positive decision about a physical movement.
An athletic coach might ask, “Is the mechanical energy of the movement better or worse than
before the new training program was instigated, and why?” Or the orthopedic surgeon may wish
to see the improvement in the knee muscle moments of a patient a month after surgery. Or a basic
researcher may wish to interpret the motor changes resulting from certain perturbations and
thereby verify or negate different theories of neural control. In all cases, if the questions asked
yield no answers, it can be said that there was no information present in the analysis. The decision
may be positive in that it may confirm that the coaching, surgery, or therapy has been correct and
should continue exactly as before. Or, if this is an initial assessment, the decision may be to pro-
ceed with a definite plan based on new information from the analysis. The information can also
cause a negative decision, for example, to cancel a planned surgical procedure and to prescribe
therapy instead.
Some biomechanical assessments involve a look at the description itself rather than some ana-
lyzed version of it. Commonly, ground reaction force curves from a force plate are examined.
This electromechanical device gives an electrical signal that is proportional to the weight (force)
of the body acting on it. Such patterns appear in Figure 1.3. A trained observer can detect pattern
changes as a result of pathological gait and may come to some conclusions as to whether the
Normal
Vertical reaction force
100%
Body weight
Pathological
Time
Figure 1.3 Example of a ground reaction force curve that has sometimes been used in the diagnostic assessment of
pathological gait.
ISTUDY
patient is improving, but he or she will not be able to assess why. At best, this approach is spec-
ulative and yields little information regarding the underlying cause of the observed patterns.
1.2 BIOMECHANICS AND ITS RELATIONSHIP WITH PHYSIOLOGY

AND ANATOMY
Because biomechanics is still an expanding discipline, it is important to identify its interaction

with other areas of movement science: neurophysiology, exercise physiology, and anatomy. The
neuromuscular system acts to control the release of metabolic energy for the purpose of generat-
ing controlled patterns of tension at the tendon. That tension waveform is a function of the
physiological characteristics of the muscle (i.e. fiber type) and of its metabolic state (rested
vs. fatigued). The tendon tension is generated in the presence of passive anatomical structures
(ligaments, articulating surfaces, and skeletal structures). Figure 1.4 depicts the relationship
between the sensory system, the neurological pathways, the muscles, the skeletal system, and
the link-segment model that we analyze. The essential characteristic of this total system is that
it is converging in nature. The structure of the neural system has many excitatory and inhibitory
synaptic junctions, all summing their control on a final synaptic junction in the spinal cord to
control individual motor units. The α motoneuron (1), which is often described as the final com-
mon pathway, has its synapse on the motor end point of the muscle motor unit. A second level of
convergence is the summation of all twitches from all active motor units at the level of the tendon
(2). This summation results from the neural recruitment of motor units based on the size principle
(DeLuca et al., 1982; Henneman and Olson, 1965). The resultant tension is a temporal superpo-
sition of twitches of all active motor units, modulated by the length and velocity characteristics of
the muscle. A third level of musculoskeletal integration at each joint center where the moment-of-
force (3) is the algebraic summation of the force/moment products of all muscles crossing that
joint plus the moments generated by the passive anatomical structures at the joint. The moments
Neuro-musculo-skeletal integration
n
Fj(fl) = ∑ Twitchj × Ratej
j–1
α (fl) 2
Ia
Ib 2
α (ext)
m
1 Final common pathway Fi(ext) = ∑ Twitchi × Ratei
n m i–1
∑ exit. + ∑ inhib.
i–1 i–1
s m n
4 Motor synergies = ∑ Mi M = ∑ Fi(ext) × di + ∑ Fj(fl) × dj
i–1 i–1 j–1
i.e, Support moment, Ms = Mα + Mk + Mh (Final common mechanical pathway)
Figure 1.4 Four levels of integration in the neuromusculoskeletal system provide control of human movement.
The first is the neural summation of all excitatory/inhibitory inputs to the α motoneuron (1). The second is the
summation of all motor twitches from the recruitment of all active motor units within the muscle and is seen as
a tendon force (2). The third is the algebraic summation of all agonist and antagonist muscle moments at the
joint axis (3). Finally, integrations are evident in combined moments acting synergistically toward a common
goal (4).
ISTUDY
1.3 REFERENCES 7
we routinely calculate include the net summation of all agonist and antagonist muscles crossing
that joint, whether they are single- or double-joint muscles. In spite of the fact that this moment
signal has mechanical units (N m), we must consider the moment signal as a neurological signal
because it represents the final desired central nervous system (CNS) control. Finally, an interseg-
ment integration may be present when the moments at two or more joints collaborate toward a
common goal. This collaboration is called a synergy. One such synergy (4), referred to as the
support moment, quantifies the integrated activity of all muscles of the lower limb in their defense
against a gravity-induced collapse during walking (Winter and Overall, 1980; Winter, 1984).
Bernstein (1967) predicted that the CNS exerts control at the level of the joints or at the syn-
ergy level when he postulated the “principle of equal simplicity” because “it would be incredibly
complex to control each and every muscle.” One of the by-products of these many levels of inte-
gration and convergence is that there is considerably more variability at the neural (EMG) level
than at the motor level and more variability at the motor level than at the kinematic level. The
resultant variability can frustrate researchers at the neural (EMG) level, but the positive aspect of
this redundancy is that the neuromuscular system is, therefore, very adaptable (Winter, 1984).
This adaptability is very meaningful in pathological gait as a compensation for motor or skeletal
deficits. For example, a major adaptation took place in a patient who underwent a knee replace-
ment because of osteoarthritic degeneration (Winter, 1989). For years prior to the surgery, this
patient had refrained from using her quadriceps to support her during walking; the resultant
increase in bone-on-bone forces induced pain in her arthritic knee joint. She compensated by
using her hip extensors instead of her knee extensors and maintained a near-normal walking pat-
tern; these altered patterns were retained by her CNS long after the painful arthritic knee was
replaced. Therefore, this moment-of-force must be considered the final desired pattern of
CNS control, or in the case of pathological movement, it must be interpreted either as a disturbed
pattern or as a CNS adaptation to the disturbed patterns. This adaptability is discussed further in
Chapters 5, 8, and 12.
1.3 REFERENCES
Bernstein, N. A. The Coordination and Regulation of Movements. (Pergaman Press, Oxford, UK, 1967).
DeLuca, C. J., R. A. LeFever, M. P. McCue, and A. P. Xenakis. “Control Scheme Governing Concurrently
Active Motor Units During Voluntary Contractions,” J. Physiol. 329: 129–142, 1982.
Henneman, E. and C. B. Olson. “Relations between Structure and Function in the Design of Skeletal Muscle,”
J. Neurophysiol. 28: 581–598, 1965.
Winter, D. A. “Kinematic and Kinetic Patterns in Human Gait: Variability and Compensating Effects,”
Human Movement Sci. 3: 51–76, 1984.
Winter, D. A. “Biomechanics of Normal and Pathological Gait: Implications for Understanding Human Loco-
motor Control,” J. Mot. Behav. 21: 337–355, 1989.
Winter, D. and A. Overall. “Principle of Lower Limb Support during Stance Phase of Gait,” J. Biomech.
13: 923–927, 1980.
ISTUDY
2
SIGNAL PROCESSING
JOSEPH A. ZENI1, STEPHEN J. THOMAS2, AND DAVID A. WINTERS3,†
1
2
3
2.0 INTRODUCTION
All of the biomechanical variables are time-varying, and it does not matter whether the measure is
kinematic, kinetic, or electromyographic (EMG); it must be processed like any other signal. Some
of these variables are directly measured: acceleration and force signals from transducers or EMG
from bioamplifiers. Others are a product of our analyses: moments-of-force, joint reaction forces,
mechanical energy, and power. All can benefit from further signal processing to extract cleaner or
averaged waveforms, correlated to find similarities or differences, or even transformed into the
frequency domain.
This chapter will summarize the analysis techniques associated with auto- and cross-
correlations, frequency (Fourier) analysis, and their applications to correct data record length
and sampling frequency. The theory of digital filtering is presented here; however, the specific
applications of digital filtering of kinematics appear in Chapter 3 and analog filtering of EMG
in Chapter 9. The applications of ensemble averaging of variables associated with repetitive
movements are also presented.
2.1 AUTO- AND CROSS-CORRELATION ANALYSES
Autocorrelation analyzes how well a signal is correlated with itself, between the present point in
time and past and future points in time. Cross-correlation analyses evaluate how well a given
signal is correlated with another signal over past, present, and future points in time. We are famil-
iar in statistics with the Pearson product-moment correlation. It is a measure of relationship
between two variables and allows us to determine whether a variable x increases or decreases
as the variable y increases. The strength and polarity of this relationship are given by the corre-
lation coefficient: the higher the value the stronger the relationship, while the sign indicates if
†
Deceased.

Companion website:
8
ISTUDY
2.1 AUTO- AND CROSS-CORRELATION ANALYSES 9
variables x and y are increasing and decreasing together (positive correlation) or if one is increas-
ing while the other is decreasing (negative correlation). The correlation coefficient is a normal-
ized dimensionless number varying from −1 to +1.
2.1.1 Similarity to the Pearson Correlation

Consider the formula for the Pearson product-moment correlation coefficient relating two vari-
ables, x and y:
N
1
N xi − x yi − y
i=1
r= (2.1)
sx sy
where xi and yi are the ith samples of x and y, x and y are the means of x and y, and sx and sy are the
standard deviations of x and y.
The numerator of the formula is the sum of the product of the two variables after the mean
value of each variable has been subtracted. It is easy to appreciate that if x and y are random
and unrelated then (xi − x) and (yi − y) will be scattered in the x − y plane about zero (see
Figure 2.1). These products will be +ve in quadrants 1 and 3 and –ve in quadrants 2 and 4,
and provided there are enough points their sum, r, will tend toward zero, indicating no relation-
ship between the two variables.
Now if the variables are related and tend to increase and decrease together (xi − x) and (yi − y)
will fall along a line with a positive slope in the x − y plane (see Figure 2.2). When we sum the
products in Equation (2.1), we will get a finite +ve sum, and when this sum is divided by N, we
remove the influence of the number of data points. This product will have the units of the product
of the two variables, and its magnitude will also be scaled by those units. To remove those two
factors, we divide by sxsy, which normalizes the correlation coefficient so that it is dimensionless
and lies between −1 and +1.
There is an estimation error in the correlation coefficient if we have a finite number of data
points, which is true of most all physiological samples. Therefore, the level of significance
will increase or decrease with the number of data points. The calculation for the t-statistic of
the correlation coefficient includes the number of observations (n) in the numerator and strength
of relationship in the denominator (r). As the number of observed samples increases (n) for a
Figure 2.1 Scatter diagram of variable x against variable y
ISTUDY
10 SIGNAL PROCESSING
Figure 2.2 Scatter diagram showing a positive correlation between variable x and variable y.
given r-value, there is a commensurate increase in the t-statistic and an increased likelihood
of achieving statistical significance for a given relationship. Any standard statistics textbook
includes a table of significance for the coefficient r for any given number of data points.
2.1.2 Formulae for Auto- and Cross-Correlation Coefficients

The auto- and cross-correlation coefficient is simply the Pearson product-moment correlation cal-
culated on two time series of data rather than on individual measures of data. Autocorrelation, as
the name suggests, involves correlating a time series with itself. Cross-correlation, on the other
hand, correlates two independent time series. The major difference is that a correlation of time
series data does not yield a single correlation coefficient, but rather a whole series of correlation
values. This series of values is achieved by shifting one of the series forward and backward in
time, the value of this shifting will be evident later. The magnitude (+ve or –ve) of this shifting is
decided by the user and the time series of correlations is a function of the phase shift, τ. The for-
mula for the autocorrelation of x(t) is Rxx(τ):
T
1
T x t x t + τ dt
0
Rxx τ = (2.2)
Rxx 0
where x(t) has zero mean.

The formula for the cross-correlation of x(t) and y(t) is Rxy(τ):
T
1
T x t y t + τ dt
0
Rxy τ = (2.3)
Rxx 0 Ryy 0
where x(t) and y(t) have zero means.

It is easy to see the similarities between these formulae and the formula for the Pearson prod-
uct-moment coefficient. The summation sign is replaced by
ISTUDY
we now divide by T rather than N. The denominator in these two equations, as in the Pearson
equation, normalizes the correlation to be dimensionless from −1 to +1. Also, the two time series
must have a zero mean, as was the case in the Pearson formula, when the means of x and y were
subtracted. Note that the Pearson correlation is a single coefficient, while these auto- and cross-
correlations are a series of correlation scores overtime at each value of τ.
2.1.3 Four Properties of the Autocorrelation Function

Property #1. The maximum value of Rxx(τ) is Rxx(0) which, in effect, is the mean square of x(t).
For all values of the phase shift, τ, either +ve or −ve Rxx(τ) is less than Rxx(0), which can be seen
from the following proof.
From basic mathematics we know:
T
2
x t −x t−τ dt ≥ 0
0
Expanding, we get:
T
2
xt + x t − τ 2 − 2x t x t − τ dt ≥ 0
0
T T T
2 2
x t dt + x t − τ dt − 2 x t x t − τ dt ≥ 0
0 0 0
For these integrations, τ is constant; thus, the second term is equal to the first term, and the
denominator for the autocorrelation is the same for all terms and is not shown. Thus:
Rxx 0 + Rxx 0 − 2Rxx τ ≥ 0

(2.4)
Rxx 0 − Rxx τ ≥ 0
Property #2. An autocorrelation function is an even function, which means that the function
for a −ve phase shift is a mirror image of the function for a +ve phase shift. This can be easily
derived as follows; for simplicity, we will only derive the numerator of the equation:
T
1
Rxx τ = x t x t + τ dt
T
0
Substituting t = (t − τ) and taking the derivative, we have dt = dt :
T
1
Rxx τ = x t − τ x t dt = Rxx − τ (2.5)
T
0
ISTUDY
Therefore, we have to calculate only the function for +ve phase shifts because the function is a
mirror image for −ve phase shifts.
Property #3. The autocorrelation function for a periodic signal is also periodic, but the phase
of the function is lost. Consider the autocorrelation of a sine wave; again we derive only the
numerator of the equation.
x t = E sin ωt
T
1
Rxx τ = E sin ωt E sin ω t − τ dt
T
0
Using the common trig identity: sin(a) sin(b) = 1/2(cos(a − b) − cos(a + b)), we get:
T
E2 1
Rxx τ = t cos ωt − sin 2ωt + ωτ
2T 2ω 0
2
E 1
Rxx τ = T cos ωt − 0 − sin 2ωT + ωτ − sin ωt
2T 2ω
Since T is one period of sin(ωt), ∴ sin(2ωT + ωτ) − sin(ωτ) = 0 for all τ.
E2
∴ Rxx τ = cos ωτ (2.6)
2
E2
Similarly if x(t) = E cos(ωt) also Rxx τ = cos ωτ .
2
Note that Equation (2.6) is an even function as predicted by Property #2; a plot of this Rxx(τ)
after normalization is presented in Figure 2.3.
This property is useful in detecting the presence of periodic signals buried in white noise.
White noise is defined as a signal made up of a series of random points, where there is zero
Rxx(τ)
1
–1
Figure 2.3 Autocorrelation of a sine or cosine signal. Note that this is an even function and the repetitive nature of
Rxx(τ) at the frequency of the sine and cosine wave.
ISTUDY
Rxx(τ)
1
–1
Figure 2.4 Autocorrelation of white noise. Note that Rxx(τ) = 0 at all τ 0, indicating that each data point has 0
correlation with all other data points ahead and behind it in time.
correlation between the signal at any point with the signal at any point ahead of or behind it in
time. Therefore, at any τ 0 Rxx(τ) = 0 and at τ = 0 Rxx(τ) = 1. Thus, the autocorrelation of white
noise is an impulse, as shown in Figure 2.4.
If we have a signal, s(t), with added noise, n(t), we can express x(t) = s(t) + n(t), and substitut-
ing in the numerator of Equation (2.2) we get:
Rxx τ = s t +n t s t + τ + n t + τ dt
0
T T
= s t s t + τ dt + n t s t + τ dt
0 0
T T
+ s t n t + τ dt + n t n t + τ dt
0 0
Since the signal and noise are uncorrelated, the second and third terms will = 0.
∴ Rxx τ = Rss τ + Rnn τ (2.7)
Property #4. As seen in Property #3 the frequency content of x(t) is present in Rxx(τ). The
power spectral density function is the Fourier transform of Rxx(τ); more will be said about this
in the next section on frequency analysis. However, it is sometimes valuable to use the auto-
correlation function to identify any periodicity present in x(t) or to identify the presence of an
interfering signal (e.g. hum) in our biological signal. Even if there were no periodicity in x(t),
the duration of Rxx(τ) would give an indication of the frequency spectra of x(t); lower frequen-
cies result in Rxx(τ) remaining above zero for longer phase shifts, while high frequencies tend to
zero for small phase shifts.
ISTUDY
2.1.4 Three Properties of the Cross-Correlation Function

Property #1. The cross-correlation of x(t) and y(t) is not an even function. Because the two
signals are completely different, the phase shifting in the +ve direction will not result in the same
“cross products” as shifting in the –ve direction. Thus, Rxy(τ) Rxy(−τ).
Property #2. The maximum value of Rxy(τ) is not necessarily at τ = 0. The maximum +ve or
negative peak of Rxy(τ) will occur when the two signals are most in phase or most out of phase.
For example, if x(t) is a sine wave and y(t) is a cosine wave of the same frequency at τ = 0, the
signals are 90 out of phase with each other, and the cross products over one cycle will sum to
zero. However, shifting the cosine wave forward 90 will bring the two signals into phase such
that all the cross products are +ve and Rxy(τ) = 1. Shifting the cosine wave backward 90 will
bring the two signals 180 out of phase so that all the “cross products” are −ve and Rxy(τ) = −1.
A physiological example is the measurement of transmission delays (neural or muscular) to deter-
mine the conduction velocity of the signal. Consider Figure 2.5, where the signal is stimulated
and is recorded at sites S1 and S2; the distance between the sites is d. The time delay between the
S1 and S2 is t as determined from RS1S2(τ), the cross-correlation of S1 and S2. Figure 2.5 shows a
peak at τ1 = t when S2 is shifted so that it is in phase with S1.
Property #3. The Fourier transform of the cross-correlation function is the cross-spectral
density function, which is used to calculate the coherence function, which is a measure of the
common frequencies present in the two signals. This is a valuable tool in determining the transfer
function of a system in which you cannot control the frequency content of the input signal. For
example, in determining the transfer function of a muscle with EMG as an input and force an
output, we cannot control the input frequencies (Bobet and Norman, 1990).
Stimulus S1 S2
V = d/t
RS1S2(τ)
0 τ1 = t
Figure 2.5 Cross-correlation of a neural or muscular signal recorded at two sites, S1 and S2, separated by a
distance, d. Rxy(τ) reaches a peak when S2 record is shifted τ1 = t second. Thus, the velocity of the transmission
V = d/t.
ISTUDY
2.1.5 Importance in Removing the Mean Bias from the Signal

A caution that must be heeded when cross-correlating two signals is that the mean amplitude of
the waveform (DC bias) in both signals must be removed prior to calculating Rxy(τ). Most stand-
ard programs do this without your knowledge, but if you are writing your own program, you must
do so or a major error will result. Consider x(t) = s1(t) + m1 and y(t) = s2(t) + m2, where m1 and m2
are the means of s1 and s2, respectively.
T
Rxy τ = s1 t + m1 s2 t + τ + m2 dt
0
T T
= s1 t s2 t + τ dt + m1 s2 t + τ dt
0 0
T T
+ m2 s1 t dt + m1 m2 dt
0 0
Since the signals and m1 and m2 are uncorrelated, the second and third terms will = 0.
T T
∴ Rxy τ = s1 t s2 t + τ dt + m1 m2 dt
0 0
The first term is the desired cross-correlation, but a major bias will be added by the second
term, and the peak of Rxy(τ) may be grossly exaggerated.
2.1.6 Digital Implementation of Auto- and Cross-Correlation Functions

Since data are now routinely collected and stored in a computer, the implementation of the auto-
and cross-correlation is the digital equivalent of Equations (2.2) and (2.3), shown below in Equa-
tions (2.8) and (2.9)
1 N
x n −x x n + τ −x
Nn=1
Rxx τ = (2.8)
1 N 2
x n −x
Nn=1
1 N
x n −x y n + τ −y
Nn=1
Rxy τ = (2.9)
1 N
x n −x y n −y
Nn=1
Both auto- and cross-correlations are calculated for various phase shifts that a priori must be
specified by the user, and this will have an impact on the number of data points used in the for-
mulae. If, for example, x(n) and y(n) are 1000 data points, and it is desired that τ = ±100, then we
can only get 800 cross products and, therefore, N will be set to 800. Sometimes the signals of
ISTUDY
interest are periodic (such as gait); then, we can wrap the signal on itself and calculate the corre-
lations using all the data points. Such an analysis is known as a circular correlation.
2.1.7 Application of Autocorrelations

As indicated in Property #3, an autocorrelation indicates the frequency content of x(t). Figure 2.6
presents an EMG record and its autocorrelation. The upper trace (a) is the raw EMG signal, which
(a)
500
400
300
200
100
EMG (μV)
–100
–200
–300
–400
–500
0 0.1 0.2 0.3 0.4 0.5
Time (s)
(b)
1
0.8
0.6
0.4
Rxx (τ)
0.2
0
–100 –80 –60 –40 –20 0 20 40 60 80 100
–0.2
–0.4
τ (ms)
Figure 2.6 (a) is a surface EMG signal recorded for 0.5 second that does not show the presence of any 60 Hz hum
pickup. (b) is the autocorrelation of this EMG over a τ = ±100 ms. Again, note that this is an even function and
observe the presence of a periodic component closely resembling a sinusoidal wave with peaks equal to the
period of a 60 Hz (approximately 17 ms).
ISTUDY
does not show any visible evidence of hum, but the autocorrelation seen in the lower trace (b) is
an even function as predicted by Property #2 and shows the presence of 60 Hz hum. This con-
sistent electrical hum is a common source of interference in electrophysiological experiments,
including those that use EMG and electroencephalograms. Note from Figure 2.3 that Rxx(τ)
for a sinusoidal wave has its first zero crossing at ¼ of a cycle of the sinusoidal frequency; thus,
we can use that first zero crossing of Rxx(τ) to estimate the average frequency in the EMG. The
first zero crossing for this Rxx(τ) occurred at about 3 ms, representing an average period of 12 ms,
or an average frequency of about 83 Hz.
2.1.8 Applications of Cross-Correlations

2.1.8.1 Quantification of Cross-Talk in Surface Electromyography. Cross-correlations
quantify what is in common in the profiles of x(t) and y(t) but also any common signal present
in both x(t) and y(t). This may be true in the recordings from surface electrodes that are close
enough to be subject to cross-talk. In brief, cross-talk occurs when the EMG signal of an active
muscle is inadvertently detected in the recording electrode of an inactive muscle. Because a
knowledge of surface recording techniques and the biophysical basis of the EMG signal is nec-
essary to understand cross-talk, the student is referred to Section 9.2.5 in Chapter 9 for a detailed
description of how Rxy(τ) has been used to quantify cross-talk.
2.1.8.2 Measurement of Delay Between Physiological Signals. Experimental research con-

ducted to find the phase advance of one EMG signal ahead of another has been used to find bal-
ance strategies in walking (Prince et al., 1994). Balance of the head and trunk during gait against
large inertial forces is achieved by the paraspinal muscles. It was noted that the head anterior/
posterior (A/P) accelerations were severely attenuated (0.48 m/s2) compared with hip accelera-
tions (1.91 m/s2), and it was important to determine how the activity of the paraspinal muscles
contributed to this reduced head acceleration. The EMG profiles at nine vertebral levels from C7
down to L4 were analyzed to find the time delays between those balance muscles. Figure 2.7
presents the ensemble average (see Section 2.3 later) of L4 and C7 muscle profiles over the stride
250
200
EMG amplitude (uV)
150 L4
100
C7
50
0
0 10 20 30 40 50 60 70 80 90 100
% of stride
Figure 2.7 Ensemble-averaged profiles over the stride period of EMG signals of paraspinal muscles at C7 and L4
levels for one of the subjects. Note the second harmonic peaks occurring during the weight acceptance periods of the
left and right feet to balance the trunk and head. The C7 amplitude is lower than the L4 amplitude because the inertial
load above C7 is considerably lower than that above L4. More important is the timing of C7 so that it is ahead of L4,
indicating that the head is balanced first, ahead of the trunk. (With
ISTUDY
C7
T2
T4
T6
T8
T10
T12
T2
Original data
L4 Exponential approx.
–80 –60 –40 –20 0

Shift (ms)
Figure 2.8 Phase shift (ms) of the activation profiles of the paraspinal muscles relative to the profile at the L4 level.
The negative shift indicates the activation was in advance of L4. The curve fit was exponential. (With Permission of
Elsevier.)
period for one subject). A cross-correlation of these two signals showed that C7 was in advance of
L4 by about 70 ms. For all 10 young adults in this study, all signals at C7, T2, T4, T6, T8, T10,
T12, and L2 were separately cross-correlated with the L4 profile. The phase shift of these signals
is presented in Figure 2.8. The earlier turn-on of the more superior paraspinal muscles indicates a
“top-down” anticipatory strategy to stabilize the head first, then the cervical level, the thoracic
level, and finally the lumbar level. This strategy resulted in a dramatic decrease in the A/P head
acceleration over the stride period compared to the A/P acceleration of the pelvis. In a subsequent
study on fit and healthy elderly (Wieman, 1991) the head/hip acceleration (%) in the elderly
(41.9%) was significantly higher (p < 0.02) compared with that of young adults (22.7%); this
indicated that the elderly had lost this “top-down” anticipatory strategy, and the paraspinal
EMG profiles bore this out.
2.1.8.3 Measurement of Synergistic and Coactivation EMG Profiles. There is considerable

information in EMG profiles regarding the action of agonist/antagonist muscle groups during any
given activity. Recently, cross-correlation techniques have been used to quantify coactivation
patterns (agonist/antagonist active at the same time) and non-coactivation patterns (agonist/
antagonist having synergistic out of phase patterns): Nelson-Wong et al. (2008) reported a study
of left and right gluteus medius patterns during a long duration standing manual task. Because
these patterns are an excellent example of motor synergies and are also related to another medial/
lateral postural strategy their details are presented in Chapter 11.
ISTUDY
2.2 FREQUENCY ANALYSIS 19
2.2 FREQUENCY ANALYSIS
2.2.1 Introduction – Time Domain vs. Frequency Domain

All the signals that we measure and analyze have a characteristic frequency content, which we
refer to as the signal spectrum; this is a plot of all the harmonics in the signal from the lowest to the
highest. The purpose of this section is to provide a conceptual background with sufficient math-
ematical derivations to help the student collect and process data and be an intelligent collector and
consumer of commercial software. Frequency domain analysis uses a powerful transform called
the Fourier transform, named after Baron Jean-Baptiste-Joseph Fourier, a French mathematician
who developed the technique in 1807.
The knowledge of the frequency spectrum of any given signal is mandatory in making deci-
sions about collection and processing of any given signal. The spectrum is a prime determinant of
the sampling rate and number of observations that are appropriate before any analog-to-digital
conversion. The spectrum also influences decisions pertaining to the frequency of data filtering
to remove undesirable noise and movement artifacts. All these factors will be discussed in the
sections to come.
2.2.2 Discrete Fourier (Harmonic) Analysis

1. Alternating Signals. An alternating signal (often called AC, for alternating current) is one
that continuously changes over time. It may be periodic or completely random, or a com-
bination of both. Also, any signal may have a DC (direct current) component, which may be
defined as the bias value about which the AC component fluctuates. Figure 2.9 shows
example signals.
2. Frequency Content. Any of these signals can also be discussed in terms of their frequency
content. A sine (or cosine) waveform is a single frequency; any other waveform can be the
sum of a number of sine and cosine waves.
Note that the Fourier transformation (see Figure 2.10) of periodic signals has discrete frequen-
cies, while nonperiodic signals have a continuous spectrum defined by its lowest frequency, f1,
and its highest frequency, f2. To analyze a periodic signal, we must express the frequency content
in multiples of the fundamental frequency f0. These higher frequencies are called harmonics. The
third harmonic is 3f0, and the tenth harmonic is 10f0. Any perfectly periodic signal can be broken
0 0 t
t
Periodic–sine wave Periodic–saw tooth
ac
component
0 0
t dc t
component
Random signal Periodic + random + dc
Figure 2.9 Time-related waveforms demonstrate the different types of signals that may be processed.
ISTUDY
Amplitude
or
sin 2πf0t power
Fourier
Amplitude t
0 transformation f0
Amplitude
T or
Fourier power
Amplitude t
0 transformation
f0 3f0 5f0
1 Amplitude
f0 =
T or
Fourier power
Amplitude t
0 transformation
f1 f2
Figure 2.10 Relationship between a signal as seen in the time domain and its equivalent in the frequency domain.
down into its harmonic components. The sum of the proper amplitudes of these harmonics is
called a Fourier series.
Thus, a given signal V(t) can be expressed as:
V t = V dc + V 1 sin ω0 t + θ1 + V 2 sin 2ω0 t + θ2 + + V n sin nω0 t + θn (2.10)
where ω0 = 2πf0, and θn is the phase angle of the nth harmonic.

For example, a square wave of amplitude V can be described by the Fourier series of odd
harmonics:
4V 1 1
V t = sin ω0 t + sin 3 ω0 t + sin 5 ω0 t + (2.11)
π 3 5
A triangular wave of duration 2t and repeating itself every T seconds is:

2 2
2Vt 1 2 2
V t = + cos ω0 t + cos 3ω0 t + (2.12)
T 2 π 3π
Several names are given to the graph showing these frequency components: spectral plots,
harmonic plots, and spectral density functions. Each shows the amplitude or power of each fre-
quency component plotted against frequency; the mathematical process to accomplish this is
called a Fourier transformation or harmonic analysis. Figure 2.10 shows plots of time-domain
signals and their equivalents in the frequency domain.
Care must be used when analyzing or interpreting the results of any harmonic analysis. Such
analyses assume that each harmonic component is present with a constant amplitude and phase
over the total analysis period. Such consistency is evident in Equation (2.10), where amplitude Vn
and phase θn are assumed constant. However, in real life, each harmonic is not constant in
either amplitude or phase. A look at the calculation of the Fourier coefficients is needed for
any signal x(t). Over the period of time T, using the discrete Fourier transform, we calculate
n harmonic coefficients.
T
2
an = x t cos nω0 t dt (2.13)
T
0
ISTUDY
T
2
bn = x t sin nω0 t dt
T
0 (2.14)
cn = a2n + b2n
an
θn = tan − 1 (2.15)
bn
It should be noted that an and bn are calculated average values over the period of time T. Thus,
the amplitude cn and the phase θn of the nth harmonic are average values as well. A certain har-
monic may be present only for part of the time T, but the computer analysis will return an average
value, assuming that it is present over the entire time. The fact that an and bn are average values
is important when we attempt to reconstitute the original signal as is demonstrated in
Section 2.2.4.5.
The digital equivalent of the Fourier transform is important to review because it gives us some
insight into the number of calculations that are necessary. In digital form, Equations (2.13) and
(2.14) for N samples during the period T:
2 N
an = xi cos nω0 i N (2.16)
N i=0
2 N
bn = xi sin nω0 i N (2.17)
N i=0
For each of the n harmonics, N calculations are necessary. The number of harmonics that can
be analyzed is from the fundamental (n = 1) up to the Nyquist frequency, which is when there
are two samples per sine or cosine wave or when n = N/2. Therefore, for N/2 harmonics, there are
N2/2 calculations necessary for each of the sine or cosine coefficients. The total number of
calculations is N2. It should be noted that the major expense in computer time is looking up
the sine and cosine values for each of the N angles.
2.2.3 Fast Fourier Transform (FFT)

The fast Fourier transform (FFT) became necessary because of the extremely large number of
calculations necessary in the Discrete Fourier Transform. As early as 1942, Danielson and Lanc-
zos introduced the Danielson-Lanczos Lemma, which showed that the Discrete Fourier Trans-
form of length N can be broken into two separate odd and even-numbered components of
length N/2 each. In a similar manner, each N/2 component can be broken into two more odd-
and even-numbered components of length N/4 each, and each of these can be broken into two
more odd and even components of length N/8 each. Thus, the basis of the FFT is a data record
that must be a power of two (2n). Therefore, if you collect data files that are not a power of two in
length, the FFT can only accept the largest 2n length file within your data file. For example, if you
collected 1000 data points, the largest file length that could be used would be 512 points; thus,
488 points would be wasted. Therefore, it is advisable to prearrange data collection files to be
binary in length; in the case of the previous example, a data file of 1024 points would be appro-
priate. With the advent of computers, many FFT algorithms appeared (Brigham, 1974), and in the
mid-1960s, J. W. Cooley and J. W. Tukey at IBM developed the most commonly used FFT algo-
rithm (Cooley and Tukey, 1965).
ISTUDY
One of the major savings in the FFT is to avoid repetitive and time-consuming calculations
especially sines and cosines. If we look at Equations (2.16) and (2.17), we see that for the fun-
damental frequency (n = 1), we must calculate N sine and N cosine values. For the second har-
monic, we recalculate every second sine and cosine value, and for the third harmonic we
recalculate every third sine and cosine value, and so on up to the highest harmonic. The FFT
calculates all sine and cosine values for the fundamental and this forms a “look-up” table for
the fundamental plus all higher harmonics. Further computational savings are achieved by clus-
tering all the products of xi and the same sine value, then summing all the xi values, and then
carrying out one product with the sine value. The number of calculations for the FFT = N log2N,
which is considerably less than N2 for the Discrete Fourier Transform. For example, for
N = 65,536 and a CPU cycle time of 1 ns (1 GHz), the DFT would take (65,5362)−9 = 4.29 s,
while the FFT would take (65536log265536)−9 = 0.001 s.
2.2.4 Applications of Spectrum Analyses

2.2.4.1 Analog-to-Digital Converters. To students not familiar with electronics, the process
that takes place during conversion of a physiological signal into a digital computer can be some-
what mystifying. A short schematic description of that process is now given. An electrical signal
representing a force, an acceleration, an EMG potential, or similar is fed into the input terminals
of the analog-to-digital converter. The computer controls the rate at which the signal is sampled;
the optimal rate is governed by the sampling theorem (see Section 2.2.4.2).
Figure 2.11 depicts the various stages in the conversion process. The first is a sample/hold
circuit in which the analog input signal is changed into a series of short-duration pulses, each
one equal in amplitude to the original analog signal at the time of sampling. The final stage
of conversion is to translate the amplitude and polarity of the sampled pulse into digital format.
This is usually a binary code in which the signal is represented by a number of bits. For example, a
16-bit code represents 216 = 65,536 levels. This means that the original sampled analog signal can
be broken into 65,536 discrete amplitude levels with a unique code representing each of these
levels. Each coded sample (consisting of zero and one second) forms a 16-bit “word,” which
is rapidly stored in computer memory for recall at a later time. If five seconds of signal were
converted at a sampling rate of 100 Hz, there would be 500 data words stored in memory to rep-
resent the original five seconds signal.
t t
Sampled
Physiological Sample signal
signal hold
circuit Digital
encoder
Sampling pulses
Computer Binary
clock data
word
To
computer
memory
Figure 2.11 Schematic diagram showing the steps involved in an analog-to-digital conversion of a physiological
signal.
ISTUDY
2.2.4.2 Deciding the Sampling Rate – The Sampling Theorem. In the processing of any
time-varying data, no matter what their source, the sampling theorem must not be violated. With-
out going into the mathematics of the sampling process, the theorem states that “the process sig-
nal must be sampled at a frequency at least twice as high as the highest frequency present in the
signal itself.” If we sample a signal at too low a frequency, we get aliasing errors. This results in
false frequencies, frequencies that were not present in the original signal, being generated in the
sample data. Figure 2.12 illustrates this effect. Both signals are being sampled at the same
interval T. Signal 1 is being sampled about ten times per cycle, while signal 2 is being sampled
less than twice per cycle. Note that the amplitudes of the samples taken from signal 2 are identical
to those sampled from signal 1. A false set of sampled data has been generated from signal 2
because the sample rate is too low – the sampling theorem has been violated.
When performing motion analysis, researchers and clinicians often use digital video to assess
joint positions or angles. There is a trade-off between using a high sampling rate (or more frames
per second) and quality of the image. At higher frames per second, there is a drop in resolution of
the images and a greater computational expense to capture the frames. This was more problematic
before the advent of digital video, but higher frame rates require greater computational processing
during analysis and greater storage space for larger files. For normal and pathological gait studies,
it has been shown that kinetic and energy analyses can be done with negligible error using a stand-
ard 24-frame per second movie camera, in part because of the frequency of joint movements and
physiological signals (accelerations, reaction forces) during walking is low (Winter, 1982).
Figure 2.13 compares the results of kinematic analysis of the foot during normal walking,
where a 50-Hz capture rate was compared with 25 Hz. The data were collected at 50 Hz, and
the acceleration of the foot was calculated using every frame of data, then reanalyzed again, using
every second frame of converted data. It can be seen that the difference between the curves is
minimal; only at the peak negative acceleration was there a noticeable difference. The final deci-
sion as to whether this error is acceptable should not rest in this curve, but in your final goal. If, for
example, the final analysis was a hip and knee torque analysis, the acceleration of the foot seg-
ment may not be too important, as is evident from another walking trial, shown in Figure 2.14.
The minor differences in no way interfere with the general pattern of joint torques over the stride
period, and the assessment of the motor patterns would be identical. Thus, for movements such as
walking or for slow movements, an inexpensive camera at 24 frames per second appears to be
quite adequate. Faster movement, such as sprinting or throwing may require higher frame rates to
adequately capture the movement without aliasing errors.
Proper
Signal 1 sampling rate
Sampling rate
Aliasing error too low
Signal 2
Sampling pulses
T
Figure 2.12 Sampling of two signals, one at a proper rate, the other at too low a rate. Signal 2 is sampled at a rate
less than twice its frequency, such that its sampled amplitudes are the same as for signal 1. This represents a violation
of the sampling theorem and results in an error called aliasing.
ISTUDY
Forward acceleration of the right foot

10.
Trial code: WP03H
50 Hz
25 Hz
5.
X acceleration (m/s/s)
0.
–5.
Heel contact-right
Heel contact-right
–10.
Tor off-right
–15.
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0 1.1
Time (s)
Figure 2.13 Comparison of the forward acceleration of the right foot during walking using the same data sampled
at 50 Hz and at 25 Hz (using data from every second frame). The major pattern is maintained with minor errors at
the peaks.
40.
Trial code: WN35H
At 25 FPS
30.
At 50 FPS
20.
10.
Torque (N.H)
0.
–10.
–20.
Heel contact-right
Heel contact-right
Toe off-right
–30.
–40.
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0 1.1 1.2
Time (s)
Figure 2.14 Comparison of the hip moment of force during level walking using the same data sampled at 50 Hz
and at 25 Hz. The residual error is quite small because the joint reaction forces dominate the inertial contributions to
the net moment of force.
ISTUDY
2.2.4.3 Deciding the Record Length. The duration of record length is decided by the lowest
frequency present in the signal. In cyclical events such as walking, cycling, or swimming the
lowest frequency is easy to determine; it is the stride frequency or how often each segment of
the body repeats itself. For example, if a patient is walking at 105 steps/minute the step frequency
is 105/60 = 1.75 steps/second = 0.875 strides/second. Thus, the fundamental frequency is 0.875
Hz. However, there are a number of noncyclical movements, which do not have a defined lowest
frequency. One such “movement” is standing either quietly or in a work-related task. In quiet
standing, we model the total body as an inverted pendulum (Gage et al., 2004), which simplifies
the total body into a single weighted-average center of mass (COM) and which can be compared
with the center of pressure (COP) measured from the force plate. Figure 2.15 presents a typical
FFT of the COP and COM in the A/P direction for a subject standing quietly for 137 seconds
(8192 samples @ 60 Hz). Note that the FFT plots the amplitude of each harmonic from
0.0073 to 1 Hz. Also note the dominant low-frequency components of both COM and COP below
0.2 Hz. The length of record must be at least a minute or longer. This long record may be com-
promised when studying patients with balance disorders because they may not be able to stand
quietly for that length of time. However, for studies on normal subjects, Carpenter et al. (2001)
found that records of at least one minute were required for acceptable reliability.
2.2.4.4 Analog and Digital Filtering of Signals – Noise and Movement Artifacts. The basic
approach can be described by analyzing the frequency spectrum of both signal and noise.
Figure 2.16a shows a schematic plot of a signal and noise spectrum. As can be seen, the signal
is assumed to occupy the lower end of the frequency spectrum and overlaps with the noise, which
is usually higher frequency. Filtering of any signal is aimed at the selective rejection, or atten-
uation, of certain frequencies. In the preceding case, the obvious filter is one that passes, unatte-
nuated, the lower-frequency signals, while at the same time attenuating the higher-frequency
noise. Such a filter, called a low-pass filter, has a frequency response as shown in
Figure 2.16b. The frequency response of the filter is the ratio of the output Xo(f) of the filter
to its input Xi(f) at each frequency present. As can be seen, the response at lower frequencies
is 1.0. This means that the input signal passes through the filter unattenuated. However, there
is a sharp transition at the cutoff frequency fc so that the signals above fc are severely attenuated.
0.3
FFT of COPx, COMx (filtered at 6 Hz)
0.25
COPx mpf = 0.124 Hz

0.2 COMx mpf = 0.0565 Hz
Amplitude (cm)
0.15
Amplitude COPx
0.1 Amplitude COMx
0.05
0
0 0.2 0.4 0.6 0.8 1
Freq. (Hz)
Figure 2.15 FFT of the COM and COP in the anterior/posterior direction of a subject standing quietly for
137 seconds. (Winter (1995).)
ISTUDY
(a)
Xi(f)
Signal
Noise
Amplitude
0 Frequency
(b)
Xi(f) Xo(f)
1.0
Filter
Half power
.707
Filter
response
Xo(f)
=
Xi(f)
0 Frequency fc
(c)
Xo(f)
Signal
Noise
0 Frequency
Figure 2.16 (a) Hypothetical frequency spectrum of a waveform consisting of a desired signal and unwanted
higher frequency noise. (b) Response of low-pass filter X0( f )/Xi( f ), introduced to attenuate the noise.
(c) Spectrum of the output waveform, obtained by multiplying the amplitude of the input by the filter response
at each frequency. Higher-frequency noise is severely attenuated, while the signal is passed with only minor
distortion in transition region around fc.
The net result of the filtering process can be seen by plotting the spectrum of the output signal
Xo(f) as seen in Figure 2.16c. Two things should be noted. First, the higher-frequency noise has
been severely reduced but not completely rejected. Second, the signal, especially in the region
where the signal and noise overlap (usually around fc) is also slightly attenuated. This results in a
slight distortion of the signal. Thus, a compromise has to be made in the selection of the cutoff
frequency. If fc is set too high, less signal distortion occurs, but too much noise is allowed to pass.
Conversely, if fc is too low, the noise is reduced drastically, but at the expense of increased signal
distortion. A sharper cutoff filter will improve matters, but at an additional expense. In digital
filtering, this means a more complex digital filter and, thus, more computer time.
(a) Hypothetical frequency spectrum of a waveform consisting of a desired signal and
unwanted higher-frequency noise.
(b) Response of low-pass filter Xo(f)/Xi(f), introduced to attenuate the noise.
(c) Spectrum of the output waveform, obtained by multiplying the amplitude of the input by
the filter response at each frequency. Higher-frequency noise is severely attenuated, while
the signal is passed with only minor distortion in
ISTUDY
1.00
Mean
2σ
Normalized harmonic amplitude

0.10
0.01
0.001
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20
Harmonic number
Figure 2.17 Harmonic content of the vertical displacement of a toe marker from 20 subjects during normal
walking. Fundamental frequency (harmonic number = 1) is normalized at 1.00. Over 99% of power is contained
below the seventh harmonic. (With Permission of Elsevier.)
The first aspect that must be assessed is what the signal spectrum is as opposed to the noise
spectrum. This can readily be done, as is seen in the harmonic analysis for 20 subjects presented
in Figure 2.17. Here is the harmonic content of the vertical displacement of the toe marker in
natural walking (Winter et al., 1974). The highest harmonics were found to be in the toe and heel
trajectories, and it was found that 99.7% of the signal power was contained in the lower seven
harmonics (below 6 Hz). Above the seventh harmonic, there was still some signal power, but it
had the characteristics of “noise.” Noise is the term used to describe components of the final sig-
nal that are not the result of the process itself (in this case, walking). Noise comes from many
sources: electronic noise in optoelectric devices, spatial precision of the scan or digitizing system,
and error in marker reconstruction during digital motion analysis. If the total effect of all these
errors is random, then the true signal will have an added random component. Usually, the random
component is high frequency, as is borne out in Figure 2.17. Here, we see evidence of higher-
frequency components extending up to the twentieth harmonic, which was the highest frequency
analyzed. The presence of the higher-frequency noise is of considerable importance when we
consider the problem of trying to calculate velocities and accelerations from the displacement
data, as will be evident in later chapters.
The theory behind digital filtering (Radar and Gold, 1967) will not be covered, but the appli-
cation of low-pass digital filtering will be described in detail. As a result of the previous discus-
sion for these data on walking, the cutoff frequency of a digital filter should be set at about 6 Hz.
The format of a recursive digital filter that processes the raw data in time domain is as follows:
X 1 nT = a0 X nT + a1 X nT − T + a2 X nT − 2T
(2.18)
+ b1 X 1 nT − T + b2 X 1 nT − 2T
where
X1 = filtered output coordinates
X = unfiltered coordinate data
nT = nth sample
(nT − T) = (n − 1)th sample
(nT − 2T) = (n − 2)th sample
a0, …, b0, … = filter coefficients.
ISTUDY
These filter coefficients a0, a1, a2, b1, and b2 are constants that depend on the type and order of
the filter, the sampling frequency, and the cutoff frequency. As can be seen, the filter output
X1(nT) is a weighted version of the immediate and past raw data plus a weighted contribution
of past filtered output. The exact equations to calculate the coefficients for a Butterworth or a
critically damped filter are as follows:
tan πf c f s
ωc = (2.19)
C
where C is the correction factor for number of passes required, to be explained shortly. For a
single pass filter C = 1.
K = 2ωc for a Butterworth filter,
or, 2ωc for a critically damped filter
K2
K 2 = ω2c , a0 = , a1 = 2a0 , a2 = a0
1 + K1 + K2
2a0
K3 = , b1 = − 2a0 + K 3
K2
b2 = 1 − 2a0 − K 3 , or b2 = 1 − a0 − a1 − a2 − b1
For example, a Butterworth-type low-pass filter of second order is to be designed to cutoff at

6 Hz using camera data taken at 60 Hz (60 frames per second). As seen in Equation (2.1), the only
thing that is required to determine these coefficients is the ratio of sampling frequency to cutoff
frequency. In this case, it is 10. The design of such a filter would yield the following coefficients:
a0 = 0 067455, a1 = 0 13491, a2 = 0 067455,

b1 = 1 14298, b2 = − 0 41280
Note that the algebraic sum of all the coefficients equals 1.0000. This gives a response of unity
over the passband. Note that the same filter coefficients could be used in many different applica-
tions, as long as the ratio fs/fc is the same. For example, an EMG signal sampled at 2000 Hz with
cutoff desired at 400 Hz would have the same coefficients as one employed for video coordinates
where the video rate was 30 Hz and cutoff was 6 Hz. The number of passes, C, in Equation (2.1) is
important when filtering kinematic data in order to eliminate the phase shift of the filtered data.
This aspect of digital filtering of kinematic data will be detailed later in Chapter 3.
2.2.4.5 Fourier Reconstitution of Original Signal. Figure 2.18 is presented to illustrate a

Fourier reconstitution of the vertical trajectory of the heel of an adult walking his or her natural
cadence. A total of nine harmonics is represented here because the addition of higher harmonics
did not improve the curve of the original data. As can be seen, the harmonic reconstitution is
visibly different from the original, sufficiently so as to cause reasonable errors in subsequent bio-
mechanical analyses. This is because each harmonic amplitude and phase values are average
values, as we cautioned about in Section 2.2.2, and this is especially true for a foot marker during
gait, which has high frequencies during swing and low frequencies during stance.
2.2.4.6 Fourier Analysis of White Noise. White noise was introduced in Figure 2.4, where an
autocorrelation showed that each point has zero correlation with any points ahead or behind it in
time. In a computer, white noise can be simulated by a random number generator. The other char-
acteristic of white noise is in the frequency domain where
ISTUDY
2.3 ENSEMBLE AVERAGING OF REPETITIVE WAVEFORMS 29
Raw Y toe motion, 9th harmonic and total

0.100
0.075
0.050
Meters
0.025
0.000
0.025
0.050
0.0 0.2 0.4 0.6 0.8 1.0 1.2
Seconds
Figure 2.18 Fourier reconstruction of the vertical trajectory of a toe marker during one walking stride. The actual
trajectory is shown by the open square, the reconstruction from the first nine harmonics is plotted with open
triangles, and the contribution of the ninth harmonic is plotted with open circles. The difference between the
actual and the reconstructed waveforms is the result of the lack of stationarity in the original signal.
the whole range of the signal, and this is similar to some of the noise apparent in kinematic data
optoelectric systems. To demonstrate this frequency spectrum characteristic we present an anal-
ysis of the FFT of white noise. Figure 2.19a is a white noise signal simulated on Excel from a
random number generator with amplitude ±1 sampled at a rate of 2048 samples/second. Thus, the
highest frequency present in this signal is the Nyquist frequency of 1024 Hz, so our FFT will
cover frequencies from 0 to 1024 Hz. An FFT of the signal in Figure 2.19a is presented in
Figure 2.19b. Note that the spectrum over that range appears “noisy,” but that is because of
the fact that the FFT is plotting the amplitude at every single Hz. A more realistic plot of this
spectrum is to carry out a moving average across the full spectrum, and this was done with a
40-point moving average, which is the “white” curve in Figure 2.19b. This moving average
approaches the theoretical constant amplitude across the full spectrum; its average amplitude
is 0.330 and ranges from 0.233 to 0.446. A longer time-domain record than the 2048 points will
result in a more constant frequency plot. Also, it should be noted that some of the individual har-
monic amplitudes are greater than one, and this is not expected from the white noise signal, which
had an amplitude of ±1. What is not shown here are the phase angles of each of the harmonics;
each has a different phase angle, so there will be many cancellations as each harmonic is added.
2.3 ENSEMBLE AVERAGING OF REPETITIVE WAVEFORMS
A large number of movements that we study are cyclical in nature and, therefore, can benefit from
a cyclical average of its many variables. Gait (walking and running) is the most common repet-
itive movement, but cycling, rowing, and lifting also benefit from such averaged profiles. Both
intra- and inter-subject averaged profiles have been reported on a wide variety of kinematic,
kinetic, and EMG variables. The major benefit of such a technique is that the averaged waveform
is more reliable and the variation about the mean gives
ISTUDY
White noise
1.2
1
0.8
0.6
0.4
Amplitude
0.2
0
–0.2
–0.4
–0.6
–0.8
–1
–1.2
0 0.2 0.4 0.6 0.8 1
Time (s)
(a)
FFT of white noise
2.5
1.5
Amplitude
0.5
0
0
31
62
93
124
155
186
217
248
279
310
341
372
403
434
465
496
527
558
589
620
651
682
713
744
775
806
837
868
899
930
961
992
1023
–0.5
Frequency (Hz)
(b)
Figure 2.19 (a) A simulated white noise signal from a random number generator with amplitude ± 1 samples at
2048 samples/second. Panel (b) is the FFT of the white noise signal at 1 Hz intervals from 0 to 1024 Hz. The “white”
line passing through this FFT plot is a 40-point moving average showing that the signal has approximately equal
power across the full spectrum of the signal.
randomness of the variable. For example, in gait, the intra-subject lower limb joint angles have
minimal variability, while the moment profiles at these same joints are quite variable. This phe-
nomenon has resulted in a covariance analysis, which can readily be done; we can calculate the
mean variance at each of the joints from these ensemble averages, from which the covariances can
be determined. As a result of those analyses, a total lower
ISTUDY
2.3 ENSEMBLE AVERAGING OF REPETITIVE WAVEFORMS 31
Hip angle during gait in patients with and without

40 hip osteoarthritis
30 Controls
Hip osteoarthritis
Sagittal plane hip angle
20
10
+ Hip flexion
0
– Hip extension
–10
–20
0 20 40 60 80 100
Percentage of stance phase of gait cycle
Figure 2.20 Sagittal plane joint angles from an individual with hip osteoarthritis (dashed line) and the ensemble
average from 20 age-matched individuals without hip pain (solid line). The gray area represents ±1 standard
deviation of the normative sample. Positive values represent hip flexion and negative values represent hip extension.
and this is reported in detail in Chapter 11. Such ensemble-averaged waveforms also form the
basis of clinical assessments, where the patient’s profile superimposed on the average for a com-
parable healthy group provides a very powerful tool in diagnosing specific movement abnorm-
alities. An example of such a clinical analysis is presented in the next section, Section 2.3.1.
2.3.1 Examples of Ensemble-Averaged Profiles

Figure 2.20 shows a typical waveform from a clinical gait study of a patient with hip osteoarthri-
tis. The sagittal plane hip joint angles during the stance phase of gait are overlaid on the averaged
profiles of 20 age-matched adults. The dark, solid line represents the ensemble average from
20 age-matched individuals who did not have lower extremity pathology, while the shaded gray
area represents ±1 standard deviation of that sample. The lighter, dashed curve represents the hip
joint angles from the individual with hip osteoarthritis. In this example, it is clear that the indi-
vidual with hip osteoarthritis has a movement profile that is different than those with pathology.
Specifically, this individual has less hip flexion at the start of the stance phase and has a lack of
hip extension during the later phases of stance. Collectively, this patient has a lower overall joint
excursion and walks with what can be considered “stiff-leg gait.” The patient’s joint angles,
alongside a normative data subset with estimates of variability, allow clinicians and researchers
to identify specific movement abnormalities and apply appropriate therapies.
In this case, the stance phase of gait has been normalized to 100% to account for temporos-
patial differences in stance time between individuals. The time base of each subject’s profile must
be altered from their individual stance time (in seconds) to a 100% stance baseline. This may
condense or expand the individual’s curve but allows for consistent averaging when the actual
timing may be variable between subjects.
2.3.2 Normalization of Time Bases to 100%

Assume that for one subject there are n samples of a given variable x over the stride period and we
wish to normalize this time-domain record to N samples
ISTUDY
this normalized curve y. Therefore, each interval of the normalized curve would be n/N samples in
duration. Assume that n = 107 samples and N = 100. At N = 1, we need the value of the variable at
the 1.07th sample; thus, by linear interpolation, y1 = 0.93x1 + 0.07x2; for N = 2 we need the value
at the 2.14th sample, or y2 = 0.86x2 + 0.14x; for N = 3 we need the value at the 3.21th sample, or
y3 = 0.79x3 + 0.21x4, and so on; for N = 99 we need the value at the 105.93th sample, or
y99 = 0.07x105 + 0.93x106; and finally y100 = x107.
It is important to note in clinical studies, normalization to 100% of a gait cycle can have
unintended consequences. Removal of the factor of time (in seconds) from physiological data
precludes evaluations of measures in which time is a relevant factor. For example, in
weight-bearing joints of the lower extremity, the rate and total duration of load are thought to
impact the integrity of the biological tissues. Calculations of loading rate or total load should
not be performed on time-normalized curves.
2.3.3 Measure of Average Variability about the Mean Waveform

The most common descriptive statistical measure of data where a single measure is taken is the
coefficient of variation, CV = σ/X where σ is the standard deviation and X is the sample mean.
The CV is a variability to mean ratio, and with the ensemble average, we wish to calculate a
similar variability score; this score is the average σ over the stride period of N points divided
by the average signal (absolute value). Some researchers suggest that the rms value of the signal
replaces the absolute value in the denominator, but this changes the CV scores very little.
1 N 2
σ
N i=1 i
CV = (2.20)
1 N
Xi
N i=1
2.4 REFERENCES
Bobet, J. and R. W. Norman. “Least Squares Identification of the Dynamic Relation Between the Electromy-
ogram and Joint Moment,” J. Biomech. 23: 1275–1276, 1990.
Brigham, E. O. The Fast Fourier Transform. (Prentice-Hall, Englewood Cliffs, NJ, 1974).
Carpenter, M. G., J. S. Frank, D. A. Winter, and G. W. Paysar. “Sampling Duration Effects on Centre of
Pressure Summary Measures,” Gait Posture 13: 35–40, 2001.
Cooley, J. W. and J. W. Tukey. “An Algorithm for the Machine Calculation of Complex Fourier Series,”
Math. Comput. 19(90): 297–301, 1965.
Gage, W. G., D. A. Winter, and J. S. Frank. “Kinematic and Kinetic Validation of Inverted Pendulum Model
in Quiet Standing,” Gait Posture 19: 124–132, 2004.
Nelson-Wong, E., D. E. Gregory, D. A. Winter, and J. P. Callaghan. “Gluteus Medius Muscle Activation
Patterns as a Predictor of Low Back Pain During Standing,” Clin. Biomech. 23: 545–553, 2008.
Prince, F., D. A. Winter, P. Stergiou, and S. E. Walt. “Anticipatory Control of Upper Body Balance During
Human Locomotion,” Gait Posture 2: 19–25, 1994.
Radar, C. M. and B. Gold. “Digital Filtering Design Techniques in the Frequency Domain,” Proc. IEEE
55: 149–171, 1967.
Wieman, C. “EMG of the Trunk and Lower Limb Muscles during Gait of Elderly and Younger Subjects:
Implications for the Control of Balance,” MSc Thesis, University of Waterloo, 1991.
ISTUDY
2.4 REFERENCES 33
Winter, D. A. The Biomechanics and Motor Control of Human Gait: Normal, Elderly and Pathological,
2nd edition. (Waterloo Biomechanics, Waterloo, Ont., 1995).
Winter, D. A. “Camera Speeds for Normal and Pathological Gait Analysis,” Med. Biol. Eng. Comput.
20: 408–412, 1982.
Winter, D. A., H. G. Sidwall, and D. A. Hobson. “Measurement and Reduction of Noise in Kinematics of
Locomotion,” J. Biomech. 7: 157–159, 1974.
ISTUDY
3
KINEMATICS
AMY L. LENZ
Department of Orthopaedics, School of Medicine, University of Utah, Salt Lake City, UT, USA
3.0 HISTORICAL DEVELOPMENT AND COMPLEXITY OF PROBLEM
Interest in the actual patterns of movement of humans and animals goes back to prehistoric
times and was depicted in cave drawings, statues, and paintings. Such replications were sub-
jective impressions of the artist. It was not until a little over a century ago that the first
motion picture cameras recorded locomotion patterns of both humans and animals. Marey,
the French physiologist, used a photographic “gun” in 1885 to record displacements in
human gait and chronophotographic equipment to get a stick diagram of a runner. About
the same time, Muybridge in the United States triggered 24 cameras sequentially to record
the patterns of a running man. Since then the technology has rapidly progressed, and we
now can record and analyze everything from the gait of a child with cerebral palsy to
the performance of an elite athlete.
The term used for these descriptions of human movement is kinematics. Kinematics is not
concerned with the forces, either internal or external, that cause the movement but rather with
the details of the movement itself. A complete and accurate quantitative description of the
simplest movement requires a huge volume of data and a large number of calculations, resulting
in an enormous number of graphic plots. For example, to describe the movement of the lower
limb in the sagittal plane during one stride can require up to 50 variables. These include linear
and angular displacements, velocities, and accelerations. It should be understood that any given
analysis may use only a small fraction of the available kinematic variables. An assessment of a
running broad jump, for example, may require only the velocity and height of the body’s center of
mass. On the other hand, a mechanical power analysis of an amputee’s gait may require almost all
the kinematic variables that are available. Therefore, the importance of each kinematic variable
and interpretation of the data for each specific patient is commonly thought of as the
intersection of science and art.

Companion website:
34
ISTUDY
3.1 KINEMATIC CONVENTIONS 35
3.1 KINEMATIC CONVENTIONS
In order to keep track of all the kinematic variables (Tables B.1 and B.2), it is important to
establish a convention system. In the anatomical literature, a definite convention has been estab-
lished, and we can completely describe a movement using terms such as proximal, flexion, and
anterior. It should be noted that these terms are all relative to the anatomic position and describe
the position of one limb relative to another. They do not give us any idea as to where we are in
space. Thus, if we wish to analyze movement relative to the ground or the direction of gravity, we
must establish an absolute spatial reference system. Such conventions are mandatory when
imaging devices are used to record the movement.
3.1.1 Absolute Spatial Reference System

Several spatial reference systems have been proposed. The one utilized throughout the text is the
one often used for human gait. The vertical direction (superior-inferior) is Z, the direction of
progression (anterior–posterior) is X, and the sideways direction (medial–lateral) is Y.
Figure 3.1 depicts this convention. The positive direction is as shown. Angles must also have
a zero reference and a positive direction. Angles in the YZ plane are measured from 0 in the
Y direction, with positive angles being counterclockwise. Similarly, in the ZX plane, angles start
at 0 in the Z direction and increase positively counterclockwise. The convention for velocities
and accelerations follows correctly if we maintain the spatial coordinate convention:
x = velocity in the X direction, positive when X is increasing
y = velocity in the Y direction, positive when Y is increasing
Sagittal
Frontal plane
plane YZ
Superior
Y Transverse
XZ plane
CM
XY Posterior
Medial
Lateral
X
Anterior
Proximal Inferior
Distal
Figure 3.1 Spatial coordinate system for all data and analyses
ISTUDY
36 KINEMATICS
z = velocity in the Z direction, positive when Z is increasing

x = acceleration in the X direction, positive when x is increasing
y = acceleration in the Y direction, positive when y is increasing
z = acceleration in the Z direction, positive when z is increasing
The same applies to angular velocities and angular accelerations. A counterclockwise angular
increase is a positive angular velocity, ω. When ω is increasing, we calculate a positive angular
acceleration, α.
3.1.2 Total Description of a Body Segment in Space

The complete kinematics of any body segment requires 15 data variables, all of which are chan-
ging with time:
1. Position (x, y, z) of segment center of mass
2. Linear velocity (x, y, z) of segment center of mass
3. Linear acceleration (x, y, z) of segment center of mass
4. Angle of segment in two planes, θxy, θyz
5. Angular velocity of segment in two planes, ωxy, ωyz
6. Angular acceleration of segment in two planes, αxy, αyz
Note that the third angle data are redundant; any segment’s direction can be completely
described in two planes. For a complete description of the total body (feet + legs + thighs +
trunk + head + upper arms + forearms and hands = 14 segments), movement in three-dimen-
sional (3D) space required 15 × 14 = 210 data variables. It is no small wonder that we have
challenges describing and analyzing more complex movements. Certain simplifications can
certainly reduce the number of variables to a manageable number. In gait, for example, the
head, arms, and trunk (HAT) are often considered to be a single segment. Therefore, the data
variables in this case (seven segments) can be reduced to 105.
3.2 DIRECT MEASUREMENT TECHNIQUES
3.2.1 Goniometers
A goniometer is a special name given to the electrical potentiometer that can be attached to meas-
ure a joint angle. One arm of the goniometer is attached to one limb segment, the other to the
adjacent limb segment, and the axis of the goniometer is aligned to the joint axis. In
Figure 3.2, you can see the fitting of the goniometer to a knee joint along with the equivalent
electrical circuit. A constant voltage E is applied across the outside terminals, and the wiper
arm moves to pick off a fraction of the total voltage. The fraction of the voltage depends on
the joint angle θ.
Thus, the voltage on the wiper arm is ν = kEθ = k1θ volts.
Note that a voltage proportional to θ requires a potentiometer whose resistance varies linearly
with θ.
A goniometer designed for clinical studies is shown fitted on a patient in Figure 3.3.
Advantages
1. A goniometer is generally inexpensive.
2. The output signal is available immediately for recording or conversion into a computer.
3. Planar rotation is recorded independently of the plane of movement of the joint.
ISTUDY
3.2 DIRECT MEASUREMENT TECHNIQUES 37
θ υ = kθ
θ
Figure 3.2 Mechanical and electrical arrangement of a goniometer located at the knee joint. Voltage output is
proportional to the joint angle.
Figure 3.3 Electrogoniometer designed to accommodate changes in the axis of rotation of the knee joint, shown
here fitted on a patient. (Reproduced by permission of Chedoke-McMaster Medical Center, Hamilton, Ont.
Canada.)
Disadvantages
1. Relative angular data are given, not absolute angles, thus severely limiting the data’s
assessment value.
2. It may require an excessive length of time to fit and align, and the alignment over fat and
muscle tissue can vary over the time of the movement.
3. If a large number are fitted, movement can be encumbered by the straps and cables.
4. More complex goniometers are required for joints that do not move as hinge joints (Finley
and Karpovich, 1964).
ISTUDY
38 KINEMATICS
3.2.2 Accelerometers
As indicated by its name, an accelerometer is a device that measures acceleration. Most accel-
erometers are nothing more than force transducers designed to measure the reaction forces asso-
ciated with a given acceleration. If the acceleration of a limb segment is a and the mass inside is
m, then the force exerted by the mass is F = ma. This force is measured by a force transducer,
usually a strain gauge or piezoresistive type. The mass is accelerated against a force transducer
that produces a signal voltage V, which is proportional to the force, and since m is known and
constant, V is also proportional to the acceleration. Today almost all accelerometers are 3D trans-
ducers. A 3D transducer is nothing more than three individual accelerometers mounted at right
angles to each other, each one then reacting to the orthogonal component acting along its axis.
Even with a 3-axis accelerometer mounted on a limb, there can be problems because of limb
rotation, as indicated in Figure 3.4. In both cases, the leg is accelerating in the same absolute
direction, as indicated by vector a. The measured acceleration component an is quite different
in each case. Thus, the accelerometer is limited to those movements whose direction in space
does not change drastically or to special contrived movements, such as horizontal flexion of
the forearm about a fixed elbow joint.
A typical electric circuit of a piezoresistive accelerometer is shown in Figure 3.5. It comprises
a half-bridge consisting of two equal resistors R1. Within the transducer, resistors Ra and Rb
an
a a
an
Figure 3.4 Two movement situations where the acceleration in space is identical but the normal components are
quite different.
+
Supply
voltage
R1
+
V
Ra Unbalanced
bridge
1 2 Balance
m V
potentiometer – a +
Rb
–
a R1
Supply
voltage
–
Figure 3.5 Electrical bridge circuit used in most force transducers and accelerometers. See text for detailed
operation.
ISTUDY
change their resistances proportionally to the acceleration acting against them. With no acceler-
ation, Ra = Rb = R1, and with the balance potentiometer properly adjusted, the voltage at terminal
1 is the same as that at terminal 2. Thus, the output voltage is V = 0. With the acceleration in the
direction shown, Rb increases and Ra decreases; thus, the voltage at terminal 1 increases. The
resultant imbalance in the bridge circuit results in voltage V, proportional to the acceleration.
Conversely, if the acceleration is upward, Rb decreases and Ra increases; the bridge unbalances
in the reverse direction, giving a signal of the opposite polarity. Thus, over the dynamic range of
the accelerometer, the signal is proportional to both the magnitude and the direction of acceler-
ation acting along the axis of the accelerometer. However, if the balance potentiometer is not
properly set, we have an unbalanced bridge and we could get a voltage-acceleration relationship
like that indicated by the dashed lines.
Advantages
1. Output signal is available immediately for recording or conversion into a computer.
Disadvantages
1. Acceleration is relative to its position on the limb segment.
2. Attaching several to the body may restrict normal movement.
3. The mass of the accelerometer may result in a movement artifact, especially in rapid move-
ments or movements involving impacts (Morris and Accelerometry-A, 1973).
3.2.3 Inertial Sensors

Inertial measurement units (IMUs) are electrical devices that commonly combine the use of 3D
accelerometers and 3D gyroscopes to measure segment orientations. Gyroscopes are sensors that
measure orientation and angular velocity relative to the earth’s gravitational pull. IMUs some-
times are built with magnetometers to measure orientation in the transverse plane but are often
subjected to ferromagnetic disturbances, which result in larger errors. Absolute orientation can be
determined with the use of one sensor if global segment kinematics are of interest. However, if
joint kinematics are desired during everyday life (in which a motion capture system would not be
appropriate) then single 3D IMUs can be attached to each body segment of interest. Since gyro-
scopes suffer from drift over time, corrections often need to be performed. There are several
methods to correct drift: (1) a reset at a known static kinematic state (using accelerometer data),
(2) with the use of fusion algorithms that integrate the gyroscope and accelerometer data and
sometimes magnetometers (the same ferromagnetic disturbances can occur) to make drift correc-
tions, or (3) the use of a high pass filter.
There are several methodological approaches to measuring 3D joint kinematics from IMUs
(McGrath et al., 2018): (1) Precise manual alignment technique: IMUs are manually positioned
on the segment to align one of the IMU axes of rotations to the axis of rotation of the joint. A static
calibration trial is then recorded and used in a weighted least squares optimization routine to
determine the IMU to segment relationship. This technique is best used in joints that function
as hinge joints and have large ranges of motion like knee flexion and extension. (2) Functional
alignment techniques: these techniques use functional calibration trials that involve prescribed
movements to determine the IMU to segment relationship. This technique can be limited if
the subject is not able to perform the functional calibration movements. This is likely to occur
when performing motion analysis on injured patients or those with disabilities. (3) Model-based
techniques: these techniques decompose the relative angular velocity between two IMU sensors
into their separate axis and position components with a formulated optimization routine. These
approaches avoid having precise alignment of sensors on segments. Principle component anal-
ysis (PCA) has been used successfully to decompose the
ISTUDY
40 KINEMATICS
sensors to estimate the joint axis and allow for the creation of two segment coordinate systems.
Euler angle rotation is then applied to the two segment coordinate systems to calculate anatomic
joint angles.
Advantages
1. Continuous collection of kinematic joint data during natural everyday life and in locations
where motion capture systems are accessible.
Disadvantages
1. Positioning and alignment of sensors.
2. Need to correct for drift.
3. Ferromagnetic disturbances of magnetometers.
4. Measurement errors with all techniques to translate IMU coordinate systems to segment
coordinate systems.
3.2.4 Special Joint Angle Measuring Systems

In the area of ergonomics, special glove systems have been developed to measure the kine-
matics of the fingers and the thumb. Figure 3.6 shows the construction of a glove transducer,
which comprises a lightweight elastic glove with sensors on the proximal two joints of each
finger and thumb plus a thumb abductor sensor. Another hand-specific design is constructed
of 11 inertial sensors positioned along each of the hand and finger segments and secured with
Fiber optic etched in the area of the joint
IR source
IR detector
Light escapes when fiber bends
Less light reaches detector

Flexion sensor area
Fiber optic cables
Interface Abduction
Glove sensors
Figure 3.6 Construction and operation of a glove transducer to measure angular displacements of the fingers.
Transducer is a loop of fiber-optic cable; the amount of light returning to the detector decreases with increased
finger flexion. Each cable is calibrated for angular displacement versus detected light intensity. (Courtesy of the
Ergonomics Laboratory, Department of Kinesiology, University of Waterloo, Waterloo, Ont. Canada.)
ISTUDY
Velcro (van den Noort et al., 2016). Each sensor contains a 3D gyroscope and 3D acceler-
ometer. Anatomic calibration steps are required to determine sensor to segment coordinate
systems, which include both static and dynamic trials for both the fingers and thumb sepa-
rately. A major use for such a system has been in the study of repetitive strain injuries (Moore
et al., 1991).
3.2.5 Electromagnetic Systems

Electromagnetic (EM) tracking systems utilize magnetic fields of known locations and
magnitudes with sensors on segments of the human body to compute 6 degrees of freedom
position and orientation throughout the EM field. Systems are comprised of EM sensors (i.e.
receivers) and EM field generators (i.e. transmitters), not requiring direct line-of-sight between
transmitters and receivers (Figure 3.7). Most EM systems utilize digitization of anatomical
landmarks that are defined relative to EM sensors’ local coordinate systems to build anatomical
frames of body segments (Urbanczyk et al., 2021). In practice, EM systems can be limited by
tethered cables and allow for a few sensors to be tracked resulting in data collection rates ranging
from 40 to 250 Hz. Older systems reduced their data collection rates when more sensors were
added, but most newer systems are able to read sensor data in parallel, thus not affecting the data
collection rate of a single sensor. Different systems utilize varying technology for field generators
which can produce either changing (AC) or quasi-static (pulsed DC) EM fields. The sensors
measure magnetic flux, defined as a component of the magnetic field passing through a specific
point in space. Notably, a sensor only measures the gradient of the EM field that represents
differences in the EM field intensity between different positions. Therefore, disturbances in
the EM field due to ferromagnetic objects and other electronic devices within the proximity
of the EM tracking system can cause functional limitations of using this technology. On average
EM tracking can achieve accuracy of 1.0 mm in good environments, but distortion in the EM field
can dramatically drop accuracy (Franz et al., 2014).
Figure 3.7 Electromagnetic system setup on a participant for upper extremity testing. One sensor is applied to each
segment to track 3D position data. In this participant, a sensor was applied to the spine, scapula, and the humerus (not
pictured), which will allow 3D kinematics for the scapulothoracic articulation and the glenohumeral joint.
ISTUDY
42 KINEMATICS
Advantages
1. Localized small sensors without the requirement for line-of-sight.
Disadvantages
1. Magnetic field distortion due to ferromagnetic objects and other electronic devices causing
interference can compromise tracking accuracy.
2. Small tracking volume and limited range relative to EM field generator.
3. Possible restricted movement from tethered cable connections for participants performing
complex motions.
4. Limited number of sensors can reduce total number of tracked body segments.
3.3 IMAGING MEASUREMENT TECHNIQUES
“A picture is worth a thousand words” holds an important message for any human observer,
including the biomechanics researcher interested in human movement. Because of the complex-
ity of most movements, the only system that can possibly capture all the data is an imaging sys-
tem. Given the additional task of describing a dynamic activity, we are further challenged by
having to capture data over an extended period of time. This necessitates taking many images
at regular intervals during the event.
There are many types of imaging systems that could be used. The discussion will describe
different types: camera, optical motion capture, optoelectric, biplane fluoroscopy, and markerless
systems. Whichever system is chosen, a lens is involved; therefore, a short review of basic optics
is given here.
3.3.1 Review of Basic Lens Optics

A simple converging lens is one that creates an inverted image in focus at a distance v from
the lens. As seen in Figure 3.8, if the lens – object distance is u, then the focal length f of
the lens is:
1 1 1
= + (3.1)
f v u
Object
v
Image u
Lens
Figure 3.8 Simple focusing lens system showing relationship between the object and image.
ISTUDY
3.3 IMAGING MEASUREMENT TECHNIQUES 43
Wide angle Normal Telephoto
f3
f2
f1
Image L1
L2
L3
f1 f2 f3
Image = Object = Object = Object
L1 L2 L3
Figure 3.9 Differences in the focal length of wide angle, normal, and telephoto lenses result in an image of the
same size.
The imaging systems used for movement studies are such that the object – lens distance is quite
large compared with the lens – image distance. Therefore,
1 1 1
≈ 0, = , or f =v (3.2)
u f v
Thus, if we know the focal length of the lens system, we can see that the image size is related
to the object size by a simple triangulation. A typical focal length is 25 mm, a wide-angle lens is
13 mm, and a telephoto lens is 150 mm. A zoom lens is just one in which the focal length is
infinitely variable over a given range. Thus, as L increases, the focal length must increase
proportionately to produce the same image size. Figure 3.9 illustrates this principle. For maxi-
mum accuracy, it is highly desirable that the image be as large as possible. Thus, it is advanta-
geous to have a zoom lens rather than a series of fixed lenses; individual adjustments can be
readily made for each movement to be studied, or even during the course of the event.
3.3.2 f-Stop Setting and Field of Focus

The amount of light entering the lens is controlled by the lens opening, which is measured by its
f-stop (f means fraction of lens aperture opening). The larger the opening, the lower the f-stop
setting. Each f-stop setting corresponds to a proportional change in the amount of light allowed
in. A lens may have the following settings: 22, 16, 11, 8, 5.6, 4, 2.8, and 2. f/22 is 1/22 of the lens
diameter, and f/11 is 1/11 of the lens diameter. Thus f/11 lets in four times the light that f/22 does.
The fractions are arranged so that each one lets in twice the light of the adjacent higher setting
(e.g. f/2.8 provides twice the light of f/4).
To keep the lighting requirements to a minimum, it is obvious that the lens should be opened as
wide as possible with a low f setting. However, problems occur with the field of focus. This is
defined as the maximum and minimum range of the object that will produce a focused image. The
lower the f setting, the narrower the range over which an object will be in focus. For example, if
we wish to photograph a movement that is to move over a range from 10 to 30 ft, we cannot
reduce the f-stop below 5.6. The range set on the lens would be about 15 ft, and everything
between 10 and 30 ft would remain in focus. The final decision regarding f-stop depends on
the shutter speed of the movie camera and the film speed.
3.3.3 Television Imaging Camera Historical Development

Historically, the major difference between television and cinematography was the fact that
television had a fixed frame rate (60 Hz in North America and 50 Hz in Europe). However, as
ISTUDY
44 KINEMATICS
we have emerged into a modern digital era, television and video frame rates have increased
dramatically to the point where our human eye cannot detect discontinuous data for human
movements. We are now more limited by resolution, pixels, and capture frame rates. In yet,
appreciating the progression and technological advancements are important to contextually
evaluate modern-day equipment options for optically capturing human kinematics.
Almost all movement analysis television systems (i.e. motion capture) were developed in
university research laboratories (Eberhart and Inman, 1951). In the late 1960s, the first reports
of television-based systems started to appear: at Delft University of Technology in The
Netherlands (Furnée, 1967, according to Woltring, 1987) and at the Twenty-First Conference
engineering in medicine and biology (EMB) in Houston, Texas (Winter et al., 1968). The first
published paper on an operational system was by Dinn et al. (1970) from the Technical University
of Nova Scotia in Halifax, Nova Scotia, Canada. It was called Computer INterface for TELivision
(CINTEL) and was developed for digitizing angiographic images at four bits (16 grey levels) to
determine the time course of left ventricular volume (Trenholm et al., 1972). It was also used for
gait studies at the University of Manitoba in Winnipeg, Manitoba, Canada, where, with higher
spatial resolution and a one-bit (black/white) conversion, the circular image of a reflective
hemispheric ping pong ball attached on anatomical landmarks was digitized (Winter et al.,
1972). With about 10 pixels within each marker image, it was possible by averaging their
coordinates to improve the spatial precision of each marker from 1 cm (distance between scan
lines of each field) to about 1 mm. The 3 M Scotch® material that was used as reflective material
has been used by most subsequent experimental and commercial systems.
Jarett et al. (1976) reported a system that detected the left edge of the image of a small reflec-
tive marker that occupied one or two scan lines. Unfortunately, the spatial precision was equal to
the scan line distance, which is about 1 cm. This system was adopted and improved by VIdeo
CONvertor (VICON) in their commercial system. Both left and right edges of the marker image
were detected, and subsequently, the detected points were curve fitted via software advanced
material analysis and simulation software (AMASS) to a circle (Macleod et al., 1990). Based
on the circle fit, the centroid was calculated. Other commercial systems, such as the one devel-
oped by the Motion Analysis Corporation, use patented edge-detection techniques (Expert
Vision). Shape recognition of the entire marker image, rather than edge detection, was used
by the Elaboratore di Immagini Televisive (ELITE) system developed in Milan, Italy.
A dedicated computer algorithm operating in real-time used a cross-correlation pattern recogni-
tion technique based on size and shape (Ferrigno and Pedotti, 1985). These systems used all grey
levels in the shape detection, thus at the time improving the spatial resolution to 1/2800 of the
viewing field. Considering the field height to be about 2.5 m, this represented a precision of about
0.9 mm.
These systems have continued to advance, now with many major competitor manufacturers,
resulting in even higher accuracy and precision systems being available. The next section will
describe current standards for optical motion capture technology.
3.3.4 Optical Motion Capture

Modern-day optical motion capture systems (i.e. passive motion capture) are quite common
across the university, clinical, and industrial settings for measuring human movement by placing
reflective spherical markers on limb segments across the body to approximate the skeletal move-
ment during activities. The applications are widely varied from using motion capture to reproduce
human movement for making animations and video games, to detecting abnormal gait when pla-
nning surgical interventions to improve functional deficits (Figure 3.10). Independent of appli-
cation, the technology is the same.
Motion capture systems consist of motion capture cameras (Figure 3.11) placed around a room
to establish a calibrated capture volume. From the ring around the camera’s lens, the light-
emitting diodes (LEDs) typically emit infrared light that is
ISTUDY
Figure 3.10 Gait assessment of a pediatric patient with a history of club foot in a clinical gait laboratory. Infrared
cameras on the ceiling and walls capture the reflected light from the spherical reflective markers mounted on both
sides of the body. Studies are performed barefoot and in shod conditions. (Courtesy of the Motion Analysis Center,
Shriners Children’s Chicago, Chicago, IL.)
Figure 3.11 Typical Vicon infrared motion capture camera rigidly mounted in a motion analysis laboratory. The
infrared lights form a ring shape around the lens and can capture data up to 2000 Hz. (Courtesy of the Motion
Capture Core Facility, University of Utah, Salt Lake City, UT.)
ISTUDY
46 KINEMATICS
worn on the individual performing movements in the capture volume. Using wavelengths in the
infrared spectrum has benefits such as reducing the influence of ambient light and using pulsed
waves to reduce motion-blur artifact. Each camera records a two-dimensional (2D) series of
images, capturing each reflective marker location over time. Each camera in the system is
assigned a local 2D plane that is used, based on the calibrated relative position and orientation
with respect to other cameras, to calculate a 3D location of each marker within the global lab-
oratory reference frame. In a well-calibrated motion capture setup, the individual marker errors
can be less than a couple of millimeters. The accuracy also increases by having additional cam-
eras in the system to provide redundant marker tracking and minimize marker dropout due to
occlusions of markers throughout more complicated activities, or while tracking multiple people
in the capture volume.
There are many different types of cameras available for motion capture, and the specific needs
will depend on the experiments being performed within the laboratory. Similar to traditional opti-
cal video cameras, those used in motion analysis have varying available frame rates, fields of
view, and resolutions. There is often an inverse relationship between the frame rate and the field
of view and resolution of the cameras, such that as the frame rate increases to capture very fast
movements, the resolution and field of view will be lowered. Rapid movements, therefore, may
require additional cameras placed closer to the object of interest.
Most cameras allow orientation in a landscape or portrait direction. To ensure the greatest
overlap of the capture volume between cameras, it is best to leave them oriented in the land-
scape direction. If it is necessary to capture portion of the volume in a higher vertical direction,
it would be more beneficial to move the cameras away from the object and place at a higher
elevation, rather than change the orientation of one or several of the cameras. Placing the cam-
eras at an elevated position with a downward orientation is beneficial for multiple reasons; it
reduces the loss of field of view into the floor and reduces the likelihood of masking a camera
view area due to the visualization of another camera across the capture volume. Elevation of the
cameras can be done by three primary means: the use of tripods, mounting to a speed rail on the
wall, or mounting to a truss system that is separate from the physical wall of the laboratory.
Speed rail and truss mountings preclude routine movement of cameras between experiments,
but these systems prevent accidental touching or movement of the cameras during data
collections.
In most camera-based motion capture systems, the markers are identified through either the
reflection of infrared light back to the source camera, or from emission of infrared light from
the markers itself, as is the case in an active marker system. In either system, the position of
the marker is identified by the light emitted or reflected from it. It is important that the cameras
recognize only the markers as light sources and ignore any other reflection in the laboratory that is
coming from a source other than the markers. If extraneous reflections are seen during calibration,
it can increase the error in the calibration process. If reflections are identified during data collec-
tion trials, the system may identify these reflections as markers, and require additional postpro-
cessing to remove unwanted “ghost” markers in the trials. Ideally, the source of these reflections
should be identified and corrected prior to the start of a data collection. Reflections can occur for
various reasons, most commonly from excess ambient light reflected off shiny objects in the lab
or floor, or from reflective clothing material that is common in athletic shoes and apparel. If the
reflections cannot be removed, the cameras must be “masked” and a region that contains the
reflection will be ignored by each camera. While this may reduce the presence of ghost markers,
it also reduces the visible area of the camera, and true markers that occupy the masked camera
space will also not be detected.
Over the years, the diameter of spherical markers that these systems could accurately track has
been dramatically reduced with systems now being able to track facial movements with 3 mm
hemispherical markers. More commonly, 9–14 mm spherical markers are used for gait analysis
and full-body evaluation of movements (Figure 3.10). A minimum of three markers needs to be
ISTUDY
placed on each segment to create a local coordinate system to later calculate joint angles (see
Sections 3.6.1 and 3.6.2). More than three markers are preferred to provide redundant tracking
of a single segment and ensure more markers are visible for postprocessing the motion data. Most
software packages include algorithms to fill gaps in the data based on surrounding marker tra-
jectories and other mathematical approximations. Also, it is important that all the markers on
a single segment are not colinear because this can lead to errors in automatic marker tracking,
as well as the lack of distinct anatomical directions to create a meaningful local coordinate system
for that limb’s segment.
Many marker-set definitions currently exist in common practice. Before using any marker-set,
new users should be trained on proper placement of markers and a reliability test should be con-
ducted to ensure errors are minimized that could be a result of multiple people placing markers in
different manners on study participants. Also, there are some limitations to marker-set definitions
because they are assuming a segment is a rigid body, not accounting for underlying complexity
(i.e. foot mechanics) or skin, muscle, and fat motion artifact. Therefore, researching a marker-set
before using it to fully understand the strengths and weaknesses is critical for a meaningful exper-
imental study design. Marker set design is described in Chapter 7.
Advantages
1. All data are presented in an absolute spatial reference system, in a plane normal to the opti-
cal axis of the camera.
2. Systems are not limited to the number of markers used.
3. Multiple individuals can be tracked at the same time.
4. Comparisons can be made between conditions such as barefoot versus shod activities.
5. Marker placement is noninvasive.
Disadvantages
1. Inconsistent marker placement can cause errors in motion tracking results.
2. Motion artifact from skin, muscle, and fat between the surface of the skin and underlying
anatomical landmarks causes errors when approximating skeletal movement, particularly
in patients with a higher body mass index (BMI).
3. Depending on the marker-set used, placement of markers and postprocessing time can be
intensive.
4. Some motion capture systems cannot be used outside in daylight due to infrared light
interferences.
3.3.5 Optoelectric Techniques

In contrast to passive motion capture cameras, optoelectric techniques (i.e. active motion capture
systems) require the participant to wear tiny infrared lights on each desired anatomical landmark
or segment. Additional digitized points with an infrared light-equipped wand can be used to mark
anatomical landmarks relative to active markers placed on body segments. The first commercial
system was developed by Northern Digital in Waterloo, Ontario, Canada. Active motion capture
systems have advanced over the years but are largely the same construct. A system includes a
single fixed unit where three cameras are rigidly mounted, as shown in Figure 3.12. The left
and right cameras are mounted to face slightly inward to establish a capture volume where
the infrared lights emitted from active markers can be seen by all three cameras. Each camera
has planar definitions, where the intersection of those three camera planes can identify a unique
3D point in space. Thus, each active marker (i.e. infrared diode) pulses and the x, y, z coordinates
in the global reference frame are recorded. The 3D accuracy for the OPTOTRAK camera
ISTUDY
48 KINEMATICS
Figure 3.12 An OPTOTRAK system with three lenses. The two outside lenses face slightly inward and define
different planes while the middle lens is in more of a horizontal plane. Marker frequency can be pulsed up to
4600 Hz. (Courtesy of the Orthopaedic Research Laboratory, University of Utah, Salt Lake City, UT.)
mounted as shown in Figure 3.12 is 0.1 mm with a 0.01 mm resolution. These mobile systems can
also be preferable when conducting research with cadavers where active markers can be placed
on a bone pin-based structure to capture individual bone motion with higher precision than is
afforded by passive motion capture systems.
Advantages
1. Active markers can pulse at higher frequencies than passive motion capture systems.
2. A single unit is easily mobile for closer more detailed movement assessment.
Disadvantages
1. Encumbrance and time to fit wired light sources (e.g. infrared diodes) can be prohibitive in
certain movements.
2. The number of active marker sources is limited.
3. Rotational movements can cause data drop out due to the direct line of sight being inter-
rupted between the active markers and the camera mount.
3.3.6 Biplane Fluoroscopy

Biplane fluoroscopy has emerged as a relatively new and viable tool to capture in vivo bone
motion during dynamic activities. The technology consists of a modified system of two fluoro-
scopes that are detached, allowing for mobility of each X-ray emitter and image intensifier to be
set up in a variety of positions depending on the region of the body that is being evaluated
(Figure 3.13). Due to the newer nature of this technology, consistent terminology across the
field is lacking and this imaging technique goes by many names (i.e. dual fluoroscopy,
biplane/biplanar videoradiography, and dual fluoroscopic imaging system).
ISTUDY
Figure 3.13 On the top, a biplane fluoroscopy setup is showing two X-ray emitters (right) and two image
intensifiers (left) positioned at floor level surrounding two force platforms to simultaneously collect ground
reaction forces with in vivo ankle kinematics. On the bottom, DSX software shows tracked solutions for the
tibia, talus, and calcaneus in 3D with the two biplane fluoroscopy images calibrated with respect to each other,
a 3D model of the bones generated from computed tomography, and digitally reconstructed radiographs aligned
on the fluoroscopic images. From this in vivo kinematics of tibiotalar and subtalar joint motion can be
calculated. (Courtesy of the Orthopaedic Research Laboratory, University of Utah, Salt Lake City, UT.)
With a system positioned, the image intensifiers are approximately 90 with respect to one
another but can also be at a wider obtuse angle depending on the desired field of view. To cal-
ibrate the system and establish a known location of the two image intensifiers, a calibration cube
with tantalum beads is placed in the center of the capture volume. The known locations of all
beads are used to establish a system x, y, z position and orientation for a global reference frame.
Two additional calibration trials are acquired: a grid is placed on the face of the image intensifiers
to account for distortion around the perimeter of the field of view, and a white trial establishes
contrast balance when nothing is in the field of view.
Participant data collection begins with a static trial which establishes the pose of bones relative
to one another. Then dynamic trials occur either by positioning the participant within the capture
volume to perform an activity or proceeding through the capture volume to evaluate various
activities that are more dynamic (Figure 3.14). Data frequencies are typically 100–200 Hz for
pulsed systems, but continuous X-ray can be used at higher capture frequencies. But an important
consideration is the radiation exposure a participant will experience during a study. It is critical to
work with a medical physicist to optimize energy settings (kVp and mA) for imaging to reduce
potential radiation. Therefore, extensive institutional review board protocols need to be approved
ISTUDY
50 KINEMATICS
Healthy control
Heel strike Early stance Mid stance Late stance Toe-off
Figure 3.14 Raw fluoroscopy images are essentially X-ray video files which can be evaluated at quasi-static points
throughout the gait cycle or evaluated as time domain series data. Here single frames are visualized from a healthy
individual performing a barefoot overground walking trial. (Courtesy of the Orthopaedic Research Laboratory,
University of Utah, Salt Lake City, UT.)
with calculated dosimetry reports for the body-specific region being studied. The room also needs
to be shielded to contain radiation scatter during experiments. An environmental health and safety
department will typically monitor radiation in the room as well as for staff conducting the studies
regularly.
In order to evaluate the in vivo kinematics, a volumetric imaging dataset (i.e. computed tomog-
raphy or magnetic resonance imaging) also needs to be acquired for each participant. Participant-
specific bone models are segmented on the scan to generate a 3D bone model. Model-based
tracking is a method of creating a digitally reconstructed radiograph (DRR) of the 3D bone model
and that DRR is projected into both calibrated views of the biplane fluoroscopy data
(Figure 3.13). The 3D position is semi-automatically manipulated until the DRR aligns over
the bone in both fluoroscopic images for every frame of the collected activity. Because this tech-
nique eliminates all error due to soft tissue artifact, typical validated measurements have reported
errors of <1 mm and <1 (Kapron et al., 2014; Wang et al., 2015; Miranda et al., 2011).
While biplane fluoroscopy has many advantages, it is also limited by a small capture volume.
To compliment the localized data collection of in vivo bone motion, systems can be synced with a
passive motion capture system to also collect full-body kinematics that are outside of the fluor-
oscopic field of view.
Advantages
1. Eliminates skin, muscle, and fat motion artifact errors by measuring underlying bone move-
ment directly.
2. Enables measurement of small articular joints not possible with skin-based motion capture
(i.e. foot mechanics).
3. Sub-millimeter and sub-degree accuracy.
4. Can be synced with passive motion capture systems to collect additional data outside of the
fluoroscopy field of view.
Disadvantages
1. Exposure to ionizing radiation due to X-ray technology and varies depending on the region
of the body being imaged.
2. Limited field of view.
3. Expensive system requiring a room with radiation shielding and protection.
4. Potential for operator-tracking errors.
ISTUDY
3.3.7 Markerless Systems

Markerless motion capture using deep learning and neural network approaches has the potential
to revolutionize the field of biomechanics. Multiple cameras are setup in any location and cali-
brated with a grid system seen by multiple cameras. Participants are encouraged to wear tight-
fitting clothes to minimize errors due to lack of movement visualization. The software
leverages computer vision techniques to address many limitations commonly seen in
marker-based motion capture systems (Section 3.3.4). Instead of tracking external markers
affixed to participants, it can track any feature of interest, approximating skeletal movement
of full-body motion and multiple individuals at once. Comparisons have been made to
marker-based motion capture demonstrating joint center root mean squared offsets to be around
2.5 cm for all joints except for the hip which demonstrated higher errors (Kanko et al., 2021a).
Additionally, inter-session repeatability for markerless systems has been shown to be within 3
(Kanko et al., 2021b). While markerless tracking systems are still not as accurate as marker-
based systems, they provide added ease and ability to measure human movement in a nonres-
tricted setting.
Advantages
1. No markers are attached to the participant being evaluated.
2. Calibration is quick by using a grid pattern that is seen by multiple cameras.
3. Motion capture can be easily performed outside and in non-laboratory settings, bringing
motion capture to the activity of interest, hopefully maintaining the natural execution of
human movements being studied.
4. Data can be imported to Visual3D for postprocessing analyses.
Disadvantages
1. Rotational kinematics have higher errors than typical passive motion capture systems.
2. Resolution of the camera will greatly impact the tracking accuracy.
3. Atypical poses may not be recognized by the deep learning approach, therefore requiring
additional training of the tracking algorithm.
4. Participants need to wear tight fit clothing for optimal tracking, baggy clothing will result in
higher errors.
3.3.8 Summary of Various Kinematic Systems

Since the early years of motion analysis performed by Muybridge, technological advancements
have diversified the types of systems that exist to measure human kinematics. Each laboratory
must define its special requirements before choosing a particular system to weigh the advantages
and disadvantages afforded by each approach. Even a combination of multiple systems may
achieve the best experimental goals. A clinical gait lab may settle on passive motion capture
because of the encumbrance of optoelectric and EM systems with wires and the reduced precision
of markerless systems to evaluate rotational abnormalities for treatment planning. Ergonomic and
athletic environments may require instant or near-instant feedback to the participant or athlete,
thus dictating the need for an automated system that can be deployed on the field more easily by
using IMUs or a markerless system. Basic researchers evaluating targeted joint level kinematics
with direct high precision and accuracy measurement of bone motion may require a biplane fluo-
roscopy system. And, finally, the cost of hardware and software may be the single limiting factor
that may force a compromise as to the final decision.
ISTUDY
52 KINEMATICS
3.4 CLINICAL MEASURES OF KINEMATICS
The ability to ambulate and perform activities of daily living is essential for maintaining inde-
pendence and a high quality of life. When orthopedic or neurological disorders impair physical
function, patients will often seek clinical treatment where a physician, physical therapist, occu-
pational therapist, or athletic trainer may evaluate their range of motion, walking, and mobility. In
the clinical setting, motion analysis systems may not be easily readable as described in
Section 3.3, and the ability to monitor the patient at home for further evaluation would be highly
beneficial. Therefore, the emergence of simple 2D kinematic applications for smartphones and
sensors that can be sent home with the patient enables additional data and evaluation of function
and mobility before more aggressive treatment options are pursued. However, these types of
approaches can generate massive quantities of data that are prone to noise and may oversimplify
the complexity of a patient’s disorder.
3.4.1 2-D Kinematic Apps/Sensors

With the advancements of smartphones and smart devices, data can be collected and evaluated
from our activities of daily living with greater ease than coming to a motion analysis laboratory.
However, these approaches are typically limited to 2D estimations of movement. While not as
precise and accurate as systems in Section 3.3, they can provide insightful data in real-time.
Video annotation tools (i.e. Kinovea, Dartfish™, and Mustard) can provide slow-motion video
and overlay of segment long axes to quickly evaluate misalignment that can be corrected with
modified motion routines. Additionally, small sensors (i.e. Garmin™ and Blast Motion™) that
are synced to a smartphone can provide cadence, number of steps, speed, and acceleration to
monitor when someone has possibly fallen or has a sporting technique that can cause problems
for overuse injuries. With any of these apps and sensor systems, care should be given to ensure
repeatability and reliability of measurement by also providing patients instructions if being sent
home to collect more data during daily life.
3.4.2 Sensor-Based Systems

Particularly popular in athletics, sensor-based systems can be worn during practice or gameplay
to monitor athletes’ performance and evaluate injury risks. For example, a baseball player may
wear a compression sleeve with a lightweight sensor (i.e. PULSEthrow) to estimate the torque
placed on the ulnar collateral ligament during a bullpen throwing session, long toss program, or as
a player is recovering from an injury. Football players may have a smart helmet (i.e. NoMoDiag-
nostics) that can use head acceleration data to spot a concussion in real-time. Simple sensor appli-
cations like this can soon become more widespread in athletics to better understand injury
mechanisms and detect problems in real-time.
3.5 PROCESSING OF RAW KINEMATIC DATA
3.5.1 Nature of Unprocessed Image Data

Motion capture technologies are sampling processes. They capture the movement event for a
short period of time, after which no further changes are recorded until the next field or frame.
Playing a video clip back slowly demonstrates this phenomenon: the image jumps from one posi-
tion to the next in a distinct step rather than a continuous process. The only reason that video does
not appear to jump at normal projection speeds (60 or 120 frames/second for video) is because the
eye can retain an image for a period of about 1/15 seconds. The eye’s short-term “memory”
enables the human observer to average or smooth out the
ISTUDY
3.5 PROCESSING OF RAW KINEMATIC DATA 53
The converted coordinate data from motion capture recordings are called raw data. This means
that they contain additive noise from many sources: electronic noise in nonoptical systems and
spatial precision from optical systems. All of these will result in random errors in the converted
data. It is, therefore, essential that the raw data be smoothed, and in order to understand the tech-
niques used to smooth the data, an appreciation of harmonic (or frequency) analysis is necessary.
The theory of harmonic analyses has been covered in Section 2.1; however, there are some addi-
tional special problems with the processing of kinematic data that are now discussed.
3.5.2 Signal Versus Noise in Kinematic Data

In the study of movement, the signal may be an anatomical coordinate that changes with time. For
example, in running, the Z (vertical) coordinate of the heel will have certain frequencies that will
be higher than those associated with the vertical coordinate of the knee or trunk. Similarly, the
frequency content of all trajectories will decrease in walking compared with running. In repetitive
movements, the frequencies present will be multiples (harmonics) of the fundamental frequency
(stride frequency). When walking at 120 steps per minute (2 Hz), the stride frequency is 1 Hz.
Therefore, we can expect to find harmonics at 2, 3, 4 Hz, and so on. Normal walking has been
analyzed by digital computers, and the harmonic content of the trajectories of seven leg and foot
markers was determined (Winter et al., 1974). The highest harmonics were found to be in the toe
and heel trajectories, and it was found that 99.7% of the signal power was contained in the lower
seven harmonics (below 6 Hz). The harmonic analysis for the toe marker for 20 subjects is shown
in Figure 3.15, which is the same as Figure 2.17 and is repeated to show the noise content. Above
the seventh harmonic, there was still some signal power, but it had the characteristics of “noise.”
Noise is the term used to describe components of the final signal that are not due to the process
itself (in this case, walking). Sources of noise were noted in Section 3.4.1, and if the total effect of
all these errors is random, then the true signal will have an added random component. Usually, the
random component is high frequency, as is borne out in Figure 3.15. Here you can see evidence of
higher-frequency components extending up to the twentieth harmonic, which was the highest
frequency analyzed.
1.00
Mean
2σ
Normalized harmonic amplitude
0.10
0.01
0.001
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20
Harmonic number
Figure 3.15 Harmonic content of the vertical displacement of a toe marker from 20 subjects during normal
walking. Fundamental frequency (harmonic number = 1) is normalized at 1.00. Over 99% of the power is
contained below the seventh harmonic. (Reproduced by permission
ISTUDY
54 KINEMATICS
3.5.3 Problems of Calculating Velocities and Accelerations

The presence of this higher-frequency noise is of considerable importance when we consider the
problem of trying to calculate velocities and accelerations. Consider the process of time differ-
entiation of a signal containing additive higher-frequency noise. Suppose that the signal can be
represented by a summation of N harmonics:
N
x= X n sin nω0 t + θn (3.3)
n=1
where
ω0 = fundamental frequency
n = harmonic number
Xn = amplitude of nth harmonic
θn = phase of nth harmonic.
To get the velocity in the x-direction Vx, we differentiate with respect to time:
N
dx
Vx = = nω0 X n cos nω0 t + θn (3.4)
dt n=1
Similarly, the acceleration Ax is:
N
dV x
Ax = − nω0 2 X n sin nω0 t + θn (3.5)
dt n=1
Thus, the amplitude of each of the harmonics increases with its harmonic number; for velo-
cities, they increase linearly, and for accelerations, the increase is proportional to the square of the
harmonic number. This phenomenon is demonstrated in Figure 3.15, where the fundamental, sec-
ond, and third harmonics are shown, along with their first and second-time derivatives. Assuming
that the amplitude x of all three components is the same, we can see that the first derivative (veloc-
ity) of harmonics increases linearly with increasing frequency. The first derivative of the third
harmonic is now three times that of the fundamental. For the second time derivative, the increase
repeats itself, and the third harmonic acceleration is now nine times that of the fundamental
(Figure 3.16).
In the trajectory data for gait, x1 might be 5 cm and x20 = 0.5 mm. The twentieth harmonic
noise is hardly perceptible in the displacement plot. In the velocity calculation, the twentieth
harmonic increases 20-fold so that it is now one-fifth that of the fundamental. In the acceleration
calculation, the twentieth harmonic increases by another factor of 20 and now is four times the
magnitude of the fundamental. This effect is shown if you look ahead to Figure 3.18, which plots
the acceleration of the toe during walking. The random-looking signal is the raw data differen-
tiated twice. The smooth signal is the acceleration calculated after most of the higher-frequency
noise has been removed. Techniques to remove this higher-frequency noise are now discussed.
3.5.4 Smoothing and Curve Fitting of Data

The removal of noise can be accomplished in several ways. The aims of each technique are
basically the same. However, the results differ somewhat.
ISTUDY
x
0 t 0 t 0 t
sin ω0t sin 2ω0t sin 3ω0t
dx
dt 0 0 0
ω0 cos ω0t
2ω0 cos 2ω0t
3ω0 cos 3ω0t
d2x
dt2 0 0
–ω02 sin ω0 t
–(2ω0)2 sin 2ω0 t
–(3ω0)2 sin 3ω0 t
Figure 3.16 Relative amplitude changes as a result of the time differentiation of signals of increasing frequency.
The first derivative increases the amplitude proportional to frequency; the second derivative increases the amplitude
proportionally to frequency squared. Such a rapid increase has severe implications in calculating accelerations when
the original displacement signal has high-frequency noise present.
3.5.4.1 Curve-Fitting Techniques. The basic assumption here is that the trajectory signal has
a predetermined shape and that by fitting the assumed shape to a “best fit” with the raw noisy data,
a smooth signal will result. For example, it may be assumed that the data are a certain order
polynomial:
x t = a0 + a1 t + a2 t 2 + a3 t 3 + + an t n (3.6)
By using computer techniques, the coefficients a0,…, an can be selected to give a best fit, using
such criteria as minimum mean square error.
A second type of curve fit can be made assuming that a certain number of harmonics are pres-
ent in the signal. Reconstituting the final signal as a sum of N lowest harmonics,
N
x t = a0 + an sin nω0 t + θn (3.7)
n=1
This model has a better basis, especially in repetitive movement, while the polynomial may be
better in certain nonrepetitive movement such as broad jumping. However, there are severe
assumptions regarding the consistency (stationarity) of an and θn, as was discussed previously
in Chapter 2.
A third technique, spline curve fitting, is a modification of the polynomial technique. The
curve to be fitted is broken into sections, each section starting and ending with an inflection point,
with special fitting being done between adjacent sections. The major problem with this technique
is the error introduced by the improper selection of the
ISTUDY
56 KINEMATICS
must be determined from the noisy data and, thus, are strongly influenced by the very noise that
we are trying to eliminate.
3.5.4.2 Digital Filtering – Refiltering to Remove Phase Lag of Low-Pass Filter. The fourth
and most common technique used to attenuate the noise is digital filtering, which was introduced
in Chapter 2. Digital filtering is not a curve-fitting technique like the three discussed above but is
a noise attenuation technique based on differences in the frequency content of the signal versus
the noise. However, there are some additional problems related to the low-pass filtering of the raw
kinematic coordinates, and these are now discussed. For the sake of convenience, the formulae
necessary to calculate the five coefficients of a second-order filter are repeated here:
tan πfc fs
ωc = (3.8)
C
where C is the correction factor for number of passes required, to be explained shortly. For a
single-pass, filter C = 1.
K= 2ωc for a Butterworth filter

or, 2ωc for a critically damped filter
K2
K 2 = ω2c , a0 = , a1 = 2a0 , a2 = a0
1 + K1 + K2
2a0
K3 = , b1 = − 2a0 + K 3
K2
b2 = 1 − 2a0 − K 3 , or b2 = 1 − a0 − a1 − a2 − b1
As well as attenuating the signal, there is a phase shift of the output signal relative to the input.
For this second-order filter, there is a 90 phase lag at the cutoff frequency. This will cause a
second form distortion, called phase distortion, to the higher harmonics within the bandpass
region. Even more phase distortion will occur to those harmonics above fc, but these components
are mainly noise, and they are being severely attenuated. This phase distortion may be more seri-
ous than the amplitude distortion that occurs to the signal in the transition region. To cancel out
this phase lag, the once-filtered data was filtered again, but this time in the reverse direction of
time (Winter et al., 1974). This introduces an equal and opposite phase lead so that the net phase
shift is zero. Also, the cutoff of the filter will be twice as sharp as that for single filtering. In effect,
by this second filtering in the reverse direction, we have created a fourth-order zero-phase-shift
filter, which yields a filtered signal that is back in phase with the raw data but with most of the
noise removed.
In Figure 3.17, we see the frequency response of a second-order Butter-worth filter normalized
with respect to the cutoff frequency. Superimposed on this curve is the response of the fourth-
order zero-phase-shift filter. Thus, the new cutoff frequency is lower than that of the original sin-
gle-pass filter; in this case, it is about 80% of the original. The correction factor for each additional
pass of a Butterworth filter is C = (21/n − 1)0.25, where n is the number of passes. Thus, for a dual
pass, C = 0.802. For a critically damped filter, C = (21/2n − 1)0.5; thus, for a dual pass, C = 0.435.
This correction factor is applied to Equation (3.8) and results in the cutoff frequency for the orig-
inal single-pass filter being set higher, so that after the second pass the desired cutoff frequency is
achieved. The major difference between these two filters is a compromise in the response in the
time domain. Butterworth filters have a slight overshoot in response to step- or impulse-type
inputs, but they have a much shorter rise time. Critically damped filters have no overshoot
but suffer from a slower rise time. Because impulsive-type inputs are rarely seen in human move-
ment data, the Butterworth filter is preferred.
ISTUDY
1.0
Second-order filter
0.8 Fourth-order (zero lag)

Half-power
Response
0.6
Quarter-power
0.4
0.2
0
0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8 2.0
f/fc
Figure 3.17 Response of a second-order low-pass digital filter. Curve is normalized at 1.0 at the cutoff frequency,
fc. Because of the phase lag characteristics of the filter, a second refiltering is done in the reverse direction in time,
which results in a fourth-order zero-lag filter.
Toe marker
3200
Horizontal acceleration (cm/s2)
1600
–1600 Second order filter

fcx 6.0 Hz
Filtered data
Raw data
–3200
0.0 0.5 1.0 1.5 2.0 2.5 3.0
Time (s)
Figure 3.18 Horizontal acceleration of the toe marker during normal walking as calculated from displacement data
from television. The solid line is the acceleration based on the unprocessed “raw” data; the dotted line is that
calculated after the data has been filtered with a fourth-order zero-lag low-pass digital filter. (Reproduced by
permission from the Journal of Biomechanics.)
The application of one of these filters in smoothing raw coordinate data can now be seen by
examining the data that yielded the harmonic plot in Figure 3.15. The horizontal acceleration of
this toe marker, as calculated by finite differences from the filtered data, is plotted in Figure 3.18.
Note how repetitive the filtered acceleration is and how it passes through the “middle” of the
noisy curve, as calculated using the unfiltered data. Also, note that there is no phase lag in these
filtered data because of the dual forward and reverse filtering processes.
3.5.4.3 Choice of Cutoff Frequency – Residual Analysis. There are several ways to choo-
se the best cutoff frequency. The first is to carry out a harmonic analysis as depicted in
Figure 3.15. By analyzing the power in each of the components, a decision can be made as to
how much power to accept and how much to reject. However, such a decision assumes that
the filter is ideal and has an infinitely sharp cutoff. A better
ISTUDY
58 KINEMATICS
Residual
Signal
distortion
a b
d
c Noise passed through filter
o
f′c fc
Figure 3.19 Plot of the residual between a filtered and an unfiltered signal as a function of the filter cutoff
frequency. See text for the interpretation as to where to set the cutoff frequency of the filter.
of the difference between filtered and unfiltered signals over a wide range of cutoff frequencies
(Wells and Winter, 1980). In this way, the characteristics of the filter in the transition region are
reflected in the decision process. Figure 3.19 shows a theoretical plot of residual versus fre-
quency. The residual at any cutoff frequency is calculated as follows [see Equation (3.9)] for
a signal of N sample points in time:
1 N 2
R fc = Xi − Xi (3.9)
N i=1
where
fc = is the cutoff frequency of the fourth-order dual-pass filter
Xi = is raw data at ith sample
X i = is filtered data at the ith sample using a fourth-order zero-lag filter.
If our data contained no signal, just random noise, the residual plot would be a straight
line decreasing from an intercept at 0 Hz to an intercept on the abscissa at the Nyquist frequency
(0.5 fs). The line de represents our best estimate of that noise residual. The intercept a on the
ordinate (at 0 Hz) is nothing more than the rms value of the noise, because X i for a 0-Hz filter
is nothing more than the mean of the noise over the N samples. When the data consist of true
signal plus noise, the residual will be seen to rise above the straight (dashed) line as the cutoff
frequency is reduced. This rise above the dashed line represents the signal distortion that is taking
place as the cutoff is reduced more and more.
The final decision is where fc should be chosen. The compromise is always a bal-
ance between the signal distortion and the amount of noise allowed through. If we
decide that both should be equal, then we simply project a line horizontally from a
to intersect the residual line at b. The frequency chosen is f 1c , and at this frequency,
the signal distortion is represented by bc. This is also an estimate of the noise that is passed
through the filter. Figure 3.20 is a plot of the residual of four markers from one stride of gait data,
and both vertical and horizontal coordinates were analyzed
ISTUDY
25
Residual signal amplitude mm (RMS)
Motion
20 Marker horizontal vertical
Rib
Hip
15
Heel
Ball
10 Best fit linear regression for harmonics 11 to 20:
Residual (mm) = 1.8 –0.05 n, r = 0.79
0
0 2 4 6 8 10 12 14 16 18 20
Harmonic (n)
0 2 4 6 8 10 12 14 16 18
Frequency (Hz)
Figure 3.20 Plot of the residual of four markers from a walking trial; both vertical and horizontal displacement
data. Motion data were digitized with the camera 5 m from the subject.
seen, the straight regression line that represents the noise is essentially the same for both
coordinates on all markers. This tells us that the noise content, mainly introduced by the human
digitizing process, is the same for all markers. This regression line has an intercept of 1.8 mm,
which indicates that the rms of the noise is 1.8 mm. In this case, the motion capture camera was
5 m from the subject and the image was 2 m high by 3 m wide. Thus, the rms noise is less than one
part in 1000.
Also, we see distinct differences in the frequency content of different markers. The residual
shows the more rapidly moving markers on the heel and ball to have power up to about 6 Hz,
while the vertical displacements of the rib and hip markers were limited to about 3 Hz. Thus,
through this selection technique, we could have different cutoff frequencies specified for each
marker displacement.
3.5.4.4 Optimal Cutoff Frequency. The residual analysis technique described in the previous
section suggested the choice of a frequency where the signal distortion was equal to the residual
noise. This optimal cutoff frequency applies to displacement data only. However, this may not be
the optimum frequency for all amplitudes of signal and noise, all sampling frequencies, and all
levels of differentiation: velocities versus accelerations. Giakas and Baltzopoulos (1997) showed
that the optimal cutoff frequencies depended on noise level and whether displacements, veloci-
ties, or accelerations were being considered. Unfortunately, their reference displacement signal
was reconstituted from a harmonic analysis, and in Chapter 2 this technique was shown to have
major problems because of the lack of stationarity of each harmonics amplitude and phase.
Yu et al. (1999) carried out a detailed analysis to estimate the optimum cutoff frequency for
higher-order derivatives, especially accelerations. They found the optimum cutoff frequencies
to be somewhat higher than those estimated for the displacement residual analysis. This is not
surprising when we consider that the acceleration increases
ISTUDY
60 KINEMATICS
the higher-frequency noise in the acceleration waveform will increase far more rapidly than the
signal itself. Also, when the sampling frequency, fs, increases, the sampling period, Δt = 1/fs,
decreases, and thus the noise as calculated by finite differences increases [see Equations (3.17)
and (3.18c)]. Thus, Yu et al. (1999) estimated that the optimum cutoff frequency was not only a
function of the residual between the filtered and unfiltered data but also a function of fs. Their
estimated optimal cutoff frequency, fc,2, was:
f c,2 = 0 06f s − 0 000022f 2s + 5 95 ε (3.10)
where fs is the sampling frequency and ε is the relative mean residual between Xi and X i
[terms defined in Equation (3.9)]. These authors present example acceleration curves (see
Figure 3.4 in Yu et al., 1999) that show a reasonable match between the accelerometer data
and the filtered motion capture data, except that the lag of the filtered data suggests that a sec-
ond-order low-pass filter was used rather than the desired fourth-order zero-lag filter.
3.5.5 Comparison of Some Smoothing Techniques

It is valuable to see the effect of several different curve-fitting techniques on the same set of noisy
data. The following summary of a validation experiment, which was conducted to compare
(Pezzack et al., 1977) three commonly used techniques, illustrates the wide differences in the
calculated accelerations.
Data obtained from the horizontal movement of a lever arm about a vertical axis were
recorded in three different ways. A goniometer on the axis recorded angular position, an
accelerometer mounted at the end of the arm gave tangential acceleration and thus angular
acceleration, and motion capture data gave image information that could be compared with
the angular and acceleration records. The comparisons are given in Figure 3.21.
Figure 3.21a compares the angular position of the lever arm as it was manually moved from
rest through about 130 and back to the original position. The goniometer signal and the
lever angle as analyzed from the motion capture data are plotted and compared closely.
The only difference is that the goniometer record is somewhat noisy compared with the
motion capture data.
Figure 3.21b compares the directly recorded angular acceleration, which can be calculated by
dividing the tangential acceleration by the radius of the accelerometer from the center of rotation,
with the angular acceleration as calculated via the second derivative of the digitally filtered coor-
dinate data (Winter et al., 1974). The two curves match extremely well, and the finite-difference
acceleration exhibits less noise than the directly recorded acceleration. Figure 3.21c compares the
directly recorded acceleration with the calculated angular acceleration, using a polynomial fit on
the raw angular data. A ninth-order polynomial was fitted to the angular displacement curve to
yield the following fit:
θ t = 0 064 + 2 0t − 35t 2 + 210t 3 − 430t 4 + 400t 5

(3.11)
− 170t 6 + 25t 7 + 2 2t 8 − 0 41t 9 rad
Note that θ is in radians and t in seconds. To get the curve for angular acceleration, all we need
to do is take the second time derivative to yield:
α t = − 70 + 1260t − 5160t 2 + 8000t 3 − 5100t 4

(3.12)
+ 1050t 5 + 123t 6 − 29 5t 7 rad s2
ISTUDY
(a) (b)
2.40 Theta vs time 8.00 Alpha vs time

Analog Analog
2.00 Raw film data 6.00 Filtered finite
1.60 4.00
Theta (rad)
Alpha (rad/s/s) ×10

1.20 2.00
0.80 0
0.40 –2.00
0 –4.00
1.20 1.40 1.60 1.80 2.00 2.20 2.40 2.60 2.80 3.00
Time (s) –6.00
–8.00
1.20 1.40 1.60 1.80 2.00 2.20 2.40 2.60 2.80 3.00
Time (s)
(c) (d)
Alpha vs time
8.00 Alpha vs time 8.00 Analog
Analog Finite difference
6.00 Chebyshev 6.00
4.00 4.00
2.00 2.00
0 0
–2.00 –2.00
–4.00 –4.00
–6.00 –6.00
–8.00 –8.00
1.20 1.40 1.60 1.80 2.00 2.20 2.40 2.60 2.80 3.00 1.20 1.40 1.60 1.80 2.00 2.20 2.40 2.60 2.80 3.00
Time (s) Time (s)
Figure 3.21 Comparison of several techniques used to determine the acceleration of a movement based on motion
capture displacement data. (a) Displacement angle of a simple extension/flexion as plotted from motion capture and
goniometer data. (b) Acceleration of the movement in (a) as measured by an accelerometer and as calculated from
motion capture coordinates after digital filtering. (c) Acceleration as determined from a ninth-order polynomial fit of
the displacement data compared with the directly recorded acceleration. (d) Acceleration as determined by finite-
difference technique of the raw coordinate data compared with the accelerometer curve. (Reproduced by permission
from the Journal of Biomechanics.)
This acceleration curve, compared with the accelerometer signal, shows considerable discrep-
ancy, enough to cast doubt on the value of the polynomial fit technique. The polynomial is fitted
to the displacement data in order to get an analytic curve, which can be differentiated to yield
another smooth curve. Unfortunately, it appears that a considerably higher-order polynomial
would be required to achieve even a crude fit, and the computer time might become too
prohibitive.
Finally, in Figure 3.21d, you can see the accelerometer signal plotted against angular accel-
eration as calculated by second-order finite-difference techniques using raw coordinate data. The
plot speaks for itself – the accelerations are too noisy to mean anything.
ISTUDY
62 KINEMATICS
3.6 CALCULATION OF OTHER KINEMATIC VARIABLES
3.6.1 Limb-Segment Angles

Given the coordinate data from anatomical markers at either end of a limb segment, it is an easy
step to calculate the absolute angle of that segment in space. It is not necessary that the two
markers be at the extreme ends of the limb segment, as long as they are in line with the long-
bone axis. Figure 3.22 shows the outline of a leg with seven anatomical markers in a four-segment
three-joint system. Markers 1 and 2 define the thigh in the sagittal plane. Note that, by conven-
tion, all angles are measured in a counterclockwise direction, starting with the horizontal equal
to 0 . Thus, θ43 is the angle of the leg in space and can be calculated from:
y3 − y4
θ43 = arctan (3.13)
x3 − x4
or, in more general notation,

yj − yi
θij = arctan (3.14)
xj − xi
As has already been noted, these segment angles are absolute in the defined spatial reference
system. It is, therefore, quite easy to calculate the joint angles from the angles of the two adjacent
segments.
1
Thigh angle- θ21
θ21
θKNEE – θ21– θ43
2
+ve For flexion
3 –ve For extension
Shank angle- θ43
4 θ43
θANKLE– θ43– θ65 + 90°
5 +ve For plantarflexion
θ65 θ –ve For dorsiflexion
76
θMT–PH– θ65 – θ76
6 7
Figure 3.22 Marker location and limb and joint angles using an established convention. Limb angles in the spatial
reference system are determined using counterclockwise from the horizontal as positive. Thus, angular velocities
and accelerations are also positive in the counterclockwise direction in the plane of movement; this is essential
for consistent convention use in subsequent kinetic analyses. Conventions for joint angles (which are relative)
are subject to wide variation among researchers; thus, the convention used must be clarified.
ISTUDY
3.6 CALCULATION OF OTHER KINEMATIC VARIABLES 63
3.6.2 Joint Angles

Each joint has a convention for describing its magnitude and polarity. For example, when the
knee is fully extended, it is described as 0 flexion, and when the leg moves in a posterior direc-
tion relative to the thigh, the knee is said to be in flexion. In terms of the absolute angles described
previously,
knee angle = θk = θ21 − θ43
If θ21 > θ43, the knee is flexed; if θ21 < θ43, the knee is extended.
The convention for the ankle is slightly different in that 90 between the leg and the foot is
boundary between plantarflexion and dorsiflexion. Therefore,
ankle angle = θa = θ43 − θ65 + 90
If θa is positive, the foot is plantarflexed; if θa is negative, the foot is dorsiflexed.
3.6.3 Velocities – Linear and Angular

There can be severe problems associated with the determination of velocity and acceleration
information. For the reasons outlined, we will assume that the raw displacement data have
been suitably smoothed by digital filtering and we have a set of smoothed coordinates and
angles to operate upon. To calculate the velocity from displacement data, all that is needed
is to take the finite difference. For example, to determine the velocity in the x-direction, we
calculate Δx/Δt, where Δx = xi + 1 − xi, and Δt is the time between adjacent samples xi + 1
and xi.
The velocity calculated this way does not represent the velocity at either of the sample times.
Rather, it represents the velocity of a point in time halfway between the two samples. This can
result in errors later on when we try to relate the velocity-derived information to displacement
data, and both results do not occur at the same point in time. A way around this problem is to
calculate the velocity and accelerations on the basis of 2Δt rather than Δt. Thus, the velocity
at the ith sample is:
xi+1 − xi−1
Vxi = m s (3.15)
2Δt
Note that the velocity is at a point halfway between the two samples, as depicted in
Figure 3.23. The assumption is that the line joining xi − 1 to xi + 1 has the same slope as the line
drawn tangent to the curve at xi.
For angular velocities, the formula is the same except that we use angular data rather than dis-
placement data in Equation (3.14); the angular acceleration at the ith sample is:
θi+1 − θi−1
ωi = rad s (3.16)
2Δt
3.6.4 Accelerations – Linear and Angular

Similarly, the acceleration is:
Vxi+1 −Vxi − 1
Axi = m s2 (3.17)
2Δt
ISTUDY
64 KINEMATICS
xi + 1
Tangent
xi
xi – 1
Δt Δt
Figure 3.23 Finite-difference technique for calculating the slope of a curve at the ith sample point.
Note that Equation (3.16) requires displacement data from samples i + 2 and i − 2; thus, a total
of five successive data points go into the acceleration. An alternative and slightly better calcu-
lation of acceleration uses only three successive data coordinates and utilizes the calculated velo-
cities halfway between sample times:
xi+1 − xi
Vxi+1 2 = m s (3.18a)
Δt
xi − xi−1
Vxi−1 2 = m s (3.18b)
Δt
Substituting these “halfway” velocities into Equation (3.17) we get:
xi+ 1 − 2xi + xi −1
Axi = m s2 (3.18c)
Δt 2
For angular accelerations merely replace displacement data with angular data in Equa-
tions (3.17) or (3.18a–c).
3.7 PROBLEMS BASED ON KINEMATIC DATA
1. Tables A.1 and A.2a in Appendix A, plot the vertical (Y) displacement of the raw and fil-
tered data (in centimeters) for the greater trochanter (hip) marker for frames 1 to 30. Use a
vertical scale as large as possible so as to identify the noise content of the raw data. In a few
lines, describe the results of the smoothing by the digital filter.
2. Using filtered coordinate data (see Table A.2c), plot the vertical displacement of the heel
marker from toe off right (TOR) (frame 1) to the next TOR (frame 70).
(a) Estimate the instant of heel-off during midstance. (Hint: Consider the elastic compres-
sion and release of the shoe material when arriving at your answer.)
(b) Determine the maximum height of the heel above ground level during swing. When
does this occur during the swing phase? (Hint: Consider the lowest displacement of
the heel marker during stance as an indication of
ISTUDY
3.8 REFERENCES 65
(c) Describe the vertical heel trajectory during the latter half of swing (frames 14–27),
especially the four frames immediately prior to heel right contact (HRC).
(d) Calculate the vertical heel velocity at HRC.
(e) Calculate from the horizontal displacement data the horizontal heel velocity
at HCR.
(f) From the horizontal coordinate data of the heel during the first foot flat period (frames
35–40) and the second foot flat period (frames 102–106), estimate the stride length.
(g) If one stride period is 69 frames, estimate the forward velocity of this subject.
3. Plot the trajectory of the trunk marker (rib cage) over one stride (frames 28–97).
(a) Is the shape of this trajectory what you would expect in walking?
(b) Is there any evidence of conservation of mechanical energy over the stride period?
(That is, is potential energy being converted to kinetic energy and vice versa?)
4. Determine the vertical displacement of the toe marker (Table A.2d) when it reaches its
lowest point in late stance and compare that with the lowest point during swing, and thereby
determine how much toe clearance took place. Answer: ytoe(fr.13) = 0.0485 m,
ytoe(fr.66) = 0.0333 m, clearance = 0.0152 m = 1.52 cm.
5. From the filtered coordinate data (see Table A.2b), calculate the following and check your
answer with that listed in the appropriate listings (see Tables A.3a, A.3b, A.3c, A.4).
(a) The velocity of the knee in the X direction for frame 10.
(b) The acceleration of the knee in the X direction for frame 10.
(c) The angle of the thigh and leg in the spatial reference system for frame 30.
(d) From (c) calculate the knee angle for frame 30.
(e) The absolute angular velocity of the leg for frame 30 (use angular data, Table A.3b).
(f) Using the tabulated vertical velocities of the toe, calculate its vertical acceleration for
frames 25 and 33.
6. From the filtered coordinate data in Table A.2b and A.2c, calculate the following and check
your answer from the results tabulated in Table A.3a and A.3b.
(a) The center of mass of the foot segment for frame 80.
(b) The velocity of the center of mass of the leg for frame 70. Give the answer in both
coordinate and polar form.
3.8 REFERENCES
Dinn, D. F., D. A. Winter, and B. G. Trenholm. “CINTEL-Computer Interface for Television,” IEEE Trans.
Comput. C-19: 1091–1095, 1970.
Eberhart, H. D. and V. T. Inman. “An Evaluation of Experimental Procedures Used in a Fundamental Study of
Human Locomotion,” Ann. NY Acad. Sci. 5: 1213–1228, 1951.
Ferrigno, G. and A. Pedotti. “ELITE: A Digital Dedicated Hardware System for Movement Analysis via Real-
Time TV Signal Processing,” IEEE Trans. Biomed. Eng. 32: 943–950, 1985.
Finley, F. R. and P. V. Karpovich. “Electrogoniometric Analysis of Normal and Pathological Gaits,” Res.
Quart. 35: 379–384, 1964.
Franz, A. M., T. Haidegger, W. Birkfellner, K. Cleary, T. M. Peters, and L. Maier-Hein. “Electromagnetic
Tracking in Medicine – A Review of Technology, Validation, and Applications,” IEEE Trans. Med.
Imaging 33: 1702–1725, 2014.
Furnée, E. H. “Hybrid Instrumentation etc,” Proc. 5th Int. Conf. Med. Biol. Eng., pp. 446, 1967.
Giakas, G. and V. Baltzopoulos. “Optimal Digital Filtering Requires a Different Cutoff Frequency Strategy
for Determination of the Higher Frequency Derivatives,” J.
ISTUDY
66 KINEMATICS
Jarett, M. O., B. J. Andrews, and J. P. Paul. “A Television/Computer System for the Analysis of Human
Locomotion,” in Proc. IERE Conf. on Applications of Electronics in Medicine, Southhampton, England,
1976.
Kanko, R. M., E. Laende, W. S. Selbie, and K. J. Deluzio. “Inter-Session Repeatability of Markerless Motion
Capture Gait Kinematics,” J. Biomech. 121: 110422, 2021a.
Kanko, R. M., E. Laende, E. M. Davis, W. S. Selbie, and K. J. Deluzio. “Concurrent Assessment of Gait
Kinematics Using Marker-Bassed and Markerless Motion Capture,” J. Biomech. 127: 110665, 2021b.
Kapron, A. L., S. K. Aoki, C. L. Peters, S. A. Maas, M. J. Bey, R. Zauel, and A. E. Anderson. “Accuracy and
Feasibility of Dual Fluoroscopy and Model-Based Tracking to Quantify in vivo Hip Kinematics During
Clinical Exams,” J. Appl. Biomech. 30: 461–470, 2014.
Macleod, A., J. R. W. Morris, and M. Lyster. “Close-Range Photogrammetry Meets Machine Vision,” in
SPIE Vol. 1395, A. Gruen and E. Baltsavias, Eds. (Bellingham, WA, 1990), pp. 12–17.
McGrath, T., R. Fineman, and L. Stirling. “An Auto-Calibrating Knee Flexion-Extension Axis Estimator
Using Principal Component Analysis with Inertial Sensors,” Sensors 18: 1–17, 2018.
Miranda, D. L., J. B. Schwartz, A. C. Loomis, E. L. Brainerd, B. C. Fleming, and J. J. Crisco. “Static and
Dynamic Error of a Biplanar Videoradiography System Using Marker-Based and Markerless Tracking
Techniques,” J. Biomech. Eng. 133: 121002, 2011.
Moore, A., R. Wells, and D. Ranney. “Quantifying Exposure in Occupational Manual Tasks with Cumulative
Trauma Disorder Potential,” Ergonomics 34: 1433–1453, 1991.
Morris, J. and R. W. Accelerometry-A. “Technique for the Measurement of Human Body Movements,” J.
Biomech. 6: 729–736, 1973.
van den Noort, J. C., H. K. Kortier, N. van Beek, D. H. E. J. Veeger, and P. H. Veltink. “Measuring 3D Hand
and Finger Kinematics – A Comparison between Inertial Sensing and an Opto-Electronic Marker System,”
Plos One 13: 1–16, 2016.
Pezzack, J. C., R. W. Norman, and D. A. Winter. “An Assessment of Derivative Determining Techniques
Used for Motion Analysis,” J. Biomech. 10: 377–382, 1977.
Trenholm, B. G., D. A. Winter, D. Mymin, and E. L. Lansdown. “Computer Determination of Left Ventricular
Volume Using Videodensitometry,” Med. Biol. Eng. 10: 163–173, 1972.
Urbanczyk, C. A., A. Bonfiglio, A. H. McGregor, and A. M. J. Bull. “Comparing Optical and Electromagnetic
Tracking Systems to Facilitate Compatibility in Sports Kinematics Data,” Int. Biomech. 8: 75–84, 2021.
Wang, B., K. E. Roach, A. L. Kapron, N. M. Fiorentino, C. L. Saltzman, M. Singer, and A. E. Anderson.
“Accuracy and Feasibility of High-Speed Dual Fluoroscopy and Model-Based Tracking to Measure
in vivo Ankle Arthrokinematics,” Gait Posture 41: 888–893, 2015.
Wells, R. P. and D. A. Winter. “Assessment of Signal and Noise in the Kinematics of Normal, Pathological
and Sporting Gaits,” in Proc. 1st Conf. Cdn. Soc. Biomech., Locomotion I, London, Ont., 1980.
Winter, D. A., S. A. Malcolm, and B. G. Trenholm. “System for Real-Time Conversion of Video Images of
Physiological Events,” in Proc. 21st Conf. Engineering in Medicine & Biol., Houston, Texas, 1968.
Winter, D. A., R. K. Greenlaw, and D. A. Hobson. “Television-Computer Analysis of Kinematics of Human
Gait,” Comput. Biomed. Res. 5: 498–504, 1972.
Winter, D. A., H. G. Sidwall, and D. A. Hobson. “Measurement and Reduction of Noise in Kinematics of
Locomotion,” J. Biomech. 7: 157–159, 1974.
Woltring, H. J. “Data Acquisition and Processing Systems in Functional Movement Analysis,” Minerva
Orthop. Traumatol. 38: 703–716, 1987.
Yu, B., D. Gabriel, L. Noble, and K. An. “Estimate of Optimum Cutoff Frequency for the Butterworth
Low-Pass Digital Filter,” J. Appl. Biomech. 15: 318–329, 1999.
ISTUDY
4
ANTHROPOMETRY
1
2
3
4.0 SCOPE OF ANTHROPOMETRY IN MOVEMENT BIOMECHANICS
Anthropometry is the major branch of anthropology that studies the physical measurements of the
human body to determine differences in individuals and groups. A wide variety of physical
measurements are required to describe and differentiate the characteristics of race, sex, age,
and body type. In the past, the major emphasis of these studies has been evolutionary and
historical. However, more recently, a major impetus has come from the needs of technological
developments, especially man-machine interfaces: workspace design, cockpits, pressure suits,
armor, and so on. Most of these needs are satisfied by basic linear, area, and volume measures.
However, human movement analysis requires kinetic measures as well: masses, moments of
inertia, and their locations. There exists also a moderate body of knowledge regarding the joint
centers of rotation, the origin and insertion of muscles, the angles of pull of tendons, and the
length and cross-sectional area of muscles.
4.0.1 Segment Dimensions

The most basic body dimension is the length of the segments between each joint. These vary
with body build, sex, and racial origin. Dempster (1955) and Dempster et al. (1959) have
summarized estimates of segment lengths and joint center locations relative to anatomical
landmarks. An average set of segment lengths expressed as a percentage of body height
was prepared by Drillis and Contini (1966) and is shown in Figure 4.1. These segment
proportions serve as a good approximation in the absence of better data, preferably measured
directly from the individual.
†
Deceased

Companion website:
ISTUDY
68 ANTHROPOMETRY
0.130H
0.186H 0.146H
0.129H 0.108H
0.259H 0.520H
H
0.936H
0.174H
0.870H
0.818H
0.191H
0.630H
0.720H
0.485H
0.530H
0.377H
0.285H
0.039H
0.055H
Foot breadth 0.152H
Foot length
Figure 4.1 Body segment lengths expressed as a fraction of body height H.
4.1 DENSITY, MASS, AND INERTIAL PROPERTIES
Kinematic and kinetic analyses require data regarding mass distributions, mass centers, moments
of inertia, and the like. Some of these measures have been determined directly from cadavers;
others have utilized measured segment volumes in conjunction with density tables, and more
modern techniques use scanning systems that produce the cross-sectional image at many intervals
across the segment.
4.1.1 Whole-Body Density

The human body consists of many types of tissue, each with a different density. Cortical bone has
a specific gravity greater than 1.8, muscle tissue is just over 1.0, fat is less than 1.0, and the lungs
contain light respiratory gases. The average density of the whole body is a function of body build,
called somatotype. Drillis and Contini (1966) developed an expression for body density d as a
function of ponderal index c = h/w1/3, where w is body weight (pounds) and h is body height
(inches):
d = 0 69 + 0 0297c kg L (4.1)
The equivalent expression in metric units, where body mass is expressed in kilograms and
height in meters, is:
d = 0 69 + 0 9c kg L (4.2)
ISTUDY
4.1 DENSITY, MASS, AND INERTIAL PROPERTIES 69
It can be seen that a shorter individual with higher body weight has a lower ponderal index than
a taller individual with a lower relative body weight and, therefore, the shorter individual has a
lower body density.
Example 4.1. Using Equations (4.1) and (4.2), calculate the whole-body density of an adult
whose height is 5 10 and who weighs 170 lb.
c = h w1 3
= 70 1701 3
= 12 64
Using Equation (4.1),
d = 0 69 + 0 0297c = 0 69 + 0 0297 × 12 64 = 1 065 kg L
In metric units,
h = 70 39 4 = 1 78 m, w = 170 2 2 = 77 3 kg, and

1 3
c = 1 78 77 3 = 0 418
Using Equation (4.2),
d = 0 69 + 0 9c = 0 69 + 0 9 × 0 418 = 1 066 kg L
4.1.2 Segment Densities

Each body segment has a unique combination of bone, muscle, fat, and other tissue, and the den-
sity within a given segment is not uniform. Generally, because of the higher proportion of bone,
the density of distal segments is greater than that of proximal segments, and individual segments
increase their densities as the average body density increases. Figure 4.2 shows these trends for
six limb segments as a function of whole-body density, as calculated by Equations (4.1) or (4.2)
or as measured directly (Drillis and Contini, 1966; Contini, 1972).
4.1.3 Segment Mass and Center of Mass

The terms center of mass and center of gravity are often used interchangeably. The more gen-
eral term is center of mass (COM), while the center of gravity refers to the COM in one axis
only, that defined by the direction of gravity. In the two horizontal axes, the term COM must
be used.
As the total body mass increases, so does the mass of each individual segment. Therefore, it is
possible to express the mass of each segment as a percentage of the total body mass. Table 4.1
summarizes the compiled results of several investigators. These values are utilized throughout
this text in subsequent kinetic and energy calculations. It should be noted that other tables of
inertial properties and segment masses exist (de Leva, 1996; Dumas et al., 2007) and comparisons
of these parameters have been reported (Dumas and Wojtusch, 2018). The location of the COM is
also given as a percentage of the segment length from either the distal or the proximal end. In
cadaver studies, it is quite simple to locate the COM by simply determining the center of balance
of each segment. To calculate the COM in vivo, we need the profile of cross-sectional area and
ISTUDY
70 ANTHROPOMETRY
lbs/ft3
65 66 67 68
78
1.25
76
Density of
upper extremity segments
ds
1.20 Han 74
Density of
Segment density (kg/L)
s
lower extremity arm
segments Fore 72
1.15
Foot 70
Shank 68
1.10
s Thigh 66
r arm
1.05 Uppe
64
1.03 1.04 1.05 1.06 1.07 1.08 1.09 1.10

Body density (kg/L)
Figure 4.2 Density of limb segments as a function of average body density.
length. Figure 4.3 gives a hypothetical profile where the segment is broken into n sections, each
with its mass indicated. The total mass M of the segment is:
n
M= mi (4.3)
i=1
where mi, is the mass of the ith section.
mi = d i V i
where di density of ith section

Vi volume of ith section.
If the density d is assumed to be uniform over the segment, then mi = dVi and:
n
M=d Vi (4.4)
i=1
The COM is such that it must create the same net gravitational moment of force about any
point along the segment axis as did the original distributed mass. Consider the COM to be located
a distance x from the left edge of the segment,
n
Mx = mi xi
i=1
n
(4.5)
1
x= mi xi
M i=1
We can now represent the complex distributed mass by a single mass M located at a distance
x from one end of the segment.
ISTUDY
TABLE 4.1 Anthropometric Data
Center of Mass/ Radius of Gyration/
Segment
Segment Length Segment Length
Weight/Total
Segment Definition Body Weight Proximal Distal C of G Proximal Distal Density
Hand Wrist axis/knuckle II middle finger 0.006 M 0.506 0.494 P 0.297 0.587 0.577 M 1.16
Forearm Elbow axis/ulnar styloid 0.016 M 0.430 0.570 P 0.303 0.526 0.647 M 1.13
Upper arm Glenohumeral axis/elbow axis 0.028 M 0.436 0.564 P 0.322 0.542 0.645 M 1.07
Forearm and hand Elbow axis/ulnar styloid 0.022 M 0.682 0.318 P 0.468 0.827 0.565 P 1.14
Total arm Glenohumeral joint/ulnar styloid 0.050 M 0.530 0.470 P 0.368 0.645 0.596 P 1.11
Foot Lateral malleolus/head metatarsal II 0.0145 M 0.50 0.50 P 0.475 0.690 0.690 P 1.10
Leg Femoral condyles/medial malleolus 0.0465 M 0.433 0.567 P 0.302 0.528 0.643 M 1.09
Thigh Greater trochanter/femoral condyles 0.100 M 0.433 0.567 P 0.323 0.540 0.653 M 1.05
Foot and leg Femoral condyles/medial malleolus 0.061 M 0.606 0.394 P 0.416 0.735 0.572 P 1.09
Total leg Greater trochanter/medial malleolus 0.161 M 0.447 0.553 P 0.326 0.560 0.650 P 1.06
Head and neck C7–T1 and 1st rib/ear canal 0.081 M 1.000 — PC 0.495 0.116 — PC 1.11
Shoulder mass Sternoclavicular joint/glenohumeral — 0.712 0.288 — — — 1.04
axis
Thorax C7–T1/T12–L1 and diaphragm* 0.216 PC 0.82 0.18 — — — 0.92
Abdomen T12–L1/L4–L5* 0.139 LC 0.44 0.56 — — — —
Pelvis L4–L5/greater trochanter* 0.142 LC 0.105 0.895 — — — —
Thorax and abdomen C7–T1/L4–L5* 0.355 LC 0.63 0.37 — — — —
Abdomen and pelvis T12–L1/greater trochanter* 0.281 PC 0.27 0.73 — — — 1.01
Trunk Greater trochanter/glenohumeral joint* 0.497 M 0.50 0.50 — — — 1.03
Trunk head neck Greater trochanter/glenohumeral joint* 0.578 MC 0.66 0.34 P 0.503 0.830 0.607 M —
Head, arms, and trunk Greater trochanter/glenohumeral joint* 0.678 MC 0.626 0.374 PC 0.496 0.798 0.621 —
(HAT) PC
HAT Greater trochanter/mid rib 0.678 1.142 — 0.903 1.456 — —
NOTE: These segments are presented relative to the length between the greater trochanter and the glenohumeral joint.
Source Codes: M, Dempster via Miller and Nelson; Biomechanics of Sport, Lea and Febiger, Philadelphia, 1973. P, Dempster via Plagenhoef; Patterns of Human Motion, Prentice-Hall, Inc.
Englewood Cliffs, NJ, 1971. L, Dempster via Plagenhoef from living subjects; Patterns of Human Motion, Prentice-Hall, Inc., Englewood Cliffs, NJ, 1971. C, Calculated.
ISTUDY
72 ANTHROPOMETRY
m1 m2 m3 mn
x1
x2
x3
xn
Figure 4.3 Location of the center of mass of a body segment relative to the distributed mass.
Example 4.2. From the anthropometric data in Table 4.1 calculate the coordinates of the
COM of the foot and the thigh given the following coordinates: ankle (84.9, 11.0), metatarsal
(101.1, 1.3), greater trochanter (72.1, 92.8), and lateral femoral condyle (86.4, 54.9). From
Table 4.1, the foot COM is 0.5 of the distance from the lateral malleolus (ankle) to the metatarsal
marker. Thus, the COM of the foot is:
x = 84 9 + 101 1 2 = 93 0 cm
y = 11 0 + 1 3 2 = 6 15 cm
The thigh COM is 0.433 from the proximal end of the segment. Thus, the COM of the
thigh is:
x = 72 1 + 0 433 86 4 − 72 1 = 78 3 cm
y = 92 8 − 0 433 92 8 − 54 9 = 76 4 cm
4.1.4 Center of Mass of a Multisegment System

With each body segment in motion, the COM of the total body is continuously changing with
time. It is, therefore, necessary to recalculate it after each interval of time, and this requires
knowledge of the trajectories of the COM of each body segment. Consider at a particular point
in time a three-segment system with the centers of mass as indicated in Figure 4.4. The COM of
the total system is located at (x0, y0), and each of these coordinates can be calculated separately;
M = m1 + m2 + m3, and:
m1 x1 + m2 x2 + m3 x3
x0 = (4.6)
M
m1 y1 + m2 y2 + m3 y3
y0 = (4.7)
M
ISTUDY
m3 (x3, y3)
M (x0, y0)
m2 (x y )
2, 2
(x1, y1)
m1
Figure 4.4 Center of mass of a three-segment system relative to the centers of mass of the individual segments.
The COM of the total body is a frequently calculated variable. Its usefulness in the assessment
of human movement, however, is quite limited. Some researchers have used the time history
COM to calculate the energy changes of the total body. Such a calculation is erroneous, because
the COM does not account for energy changes related to reciprocal movements of the limb seg-
ments. Thus, the energy changes associated with the forward movement of one leg and the back-
ward movement of another will not be detected in the COM, which may remain relatively
unchanged. More about this will be said in Chapter 6. The major use of the body COM is in
the analysis of sporting events, especially jumping events, where the path of the COM is critical
to the success of the event because its trajectory is decided immediately at takeoff. Also, in studies
of body posture and balance, the COM is an essential calculation.
4.1.5 Mass Moment of Inertia and Radius of Gyration

The location of the COM of each segment is needed for an analysis of translational movement
through space. If accelerations are involved, we need to know the inertial resistance to such
movements. In the linear sense, F = ma describes the relationship between a linear force F
and the resultant linear acceleration a. In the rotational sense, M = Iα. M is the moment of force
causing the angular acceleration α. Thus, I is the constant of proportionality that measures
the ability of the segment to resist changes in angular velocity. M has units of N m, α is in
rad/s2, and I is in kg m2. The value of I depends on the point about which the rotation is taking
place and is a minimum when the rotation takes place about its COM. Consider a distributed mass
segment as in Figure 4.3. The moment of inertia about the left end is:
I = m1 x21 + m2 x22 + + mn x2n

n
(4.8)
= mi x2i
i=1
It can be seen that the mass close to the center of rotation has very little influence on I, while the
furthest mass has a considerable effect. This principle is used in industry to regulate the speed of
rotating machines: the mass of a flywheel is concentrated at the perimeter of the wheel with as
large a radius as possible. Its large moment of inertia resists changes in velocity and, therefore,
tends to keep the machine speed constant.
ISTUDY
74 ANTHROPOMETRY
x
m/2 m/2
ρ ρ
Figure 4.5 Radius of gyration of a limb segment relative to the location of the center of mass of the original
system.
Consider the moment of inertia I0 about the COM. In Figure 4.5 the mass has been broken into
two equal point masses. The location of these two equal components is at a distance ρ0 from the
center such that:
I 0 = mρ20 (4.9)
ρ0 is the radius of gyration and is such that the two equal masses shown in Figure 4.5 have the
same moment of inertia in the plane of rotation about the COM as the original distributed segment
did. Note that the COM of these two equal point masses is still the same as the original
single mass.
4.1.6 Parallel Axis Theorem

Most body segments do not rotate about their mass center but rather about the joint at either end.
In vivo measures of the moment of inertia can only be taken about a joint center. The relationship
between this moment of inertia and that about the COM is given by the parallel axis theorem.
A short proof is now given.
m m
I= x − ρ0 2 + x + ρ0 2
2 2
= mρ20 + mx2 (4.10)
= I 0 + mx2
where
I0 = moment of inertia about COM
x = distance between COM and center of rotation
m = mass of segment.
Actually, x can be any distance in either direction from the COM as long as it lies along the same
axis as I0 was calculated on.
Example 4.3
(a) A prosthetic leg has a mass of 3 kg and a COM of 20 cm from the knee joint. The radius of
gyration is 14.1 cm. Calculate I about the knee joint.
2
I 0 = mρ20 = 3 0 141 = 0 06 kg m2
I = I 0 + mx2
2
= 0 06 + 3 0 2 = 0 18
ISTUDY
(b) If the distance between the knee and hip joints is 42 cm, calculate Ih for this prothesis about
the hip joint as the amputee swings through with a locked knee.
x = distance from mass center to hip = 20 + 42 = 62 cm
I = I 0 + mx2
2
= 0 06 + 3 0 62 = 1 21 kg m2
Note that Ih is about 20 times that calculated about the COM.
4.1.7 Use of Anthropometric Tables and Kinematic Data

Using Table 4.1 in conjunction with kinematic data, we can calculate many variables needed for
kinetic energy analyses (see Chapters 5 and 6). This table gives the segment mass as a fraction of
body mass and centers of mass as a fraction of their lengths from either the proximal or the distal
end. The radius of gyration is also expressed as a fraction of the segment length about the COM,
the proximal end, and the distal end.
4.1.7.1 Calculation of Segment Masses and Centers of Mass.

Example 4.4. Calculate the mass of the foot, shank, thigh, and head/arms/trunk (HAT) and its
location from the proximal or distal end, assuming that the body mass of the subject is 80 kg.
Using the mass fractions for each segment,
Mass of foot = 0 0145 × 80 = 1 16 kg

Mass of leg = 0 0465 × 80 = 3 72 kg
Mass of thigh = 0 10 × 80 = 8 0 kg
Mass of HAT = 0 678 × 80 = 54 24 kg
Direct measures yielded the following segment lengths: foot = 0.195 m, leg = 0.435 m,
thigh = 0.410 m, HAT = 0.295 m.
COM of foot = 0 50 × 0 195 = 0 098 m between ankle
and metatarsal markers
COM of leg = 0 433 × 0 435 = 0 188 m below femoral
condyle marker
COM of thigh = 0 433 × 0 410 = 0 178 m below greater
trochanter marker
COM of HAT = 1 142 × 0 295 = 0 337 m above greater
trochanter marker
where COM stands for COM.
4.1.7.2 Calculation of Total-Body Center of Mass. The calculation of the COM of the total
body is a special case of Equations (4.6) and (4.7). For an n-segment body system, the COM in the
X direction is:
m1 x1 + m2 x2 + + mn xn
x= (4.11)
m1 + m2 + + mn
where m1 + m2 + + mn = M, the total body mass.
ISTUDY
76 ANTHROPOMETRY
It is quite normal to know the values of m1 = f1M, m2 = f2M, and so on. Therefore,
f 1 Mx1 + f 2 Mx2 + + f n Mxn

x= = f 1 x1 + f 2 x2 + + f n xn (4.12)
M
This equation is easier to use because all we require is knowledge of the fraction of total
body mass and the coordinates of each segment’s COM. These fractions are given in
Table 4.1.
It is not always possible to measure the COM of every segment, especially if it is not in full
view of the camera. In the sample data that follows, we have the kinematics of the right side of
HAT and the right limb during walking. It may be possible to simulate data for the left side of
HAT and the left limb. If we assume symmetry of gait, we can say that the trajectory of the left
limb is the same as that of the right limb, but out of phase by half a stride. Thus, if we use data for
the right limb one-half stride later in time and shift them back in space one-half a stride length, we
can simulate data for the left limb and left side of HAT.
Example 4.5. Calculate the total-body COM at a given frame 15. The time for one stride was
68 frames. Thus, the data from frame 15 become the data for the right lower limb and the right half
of HAT, and the data one-half stride (34 frames) later become those for the left side of the body.
All coordinates from frame 49 must now be shifted back in the x-direction by a step length. An
examination of the x coordinates of the heel during two successive periods of stance showed the
stride length to be 264.2 − 122.8 = 141.4 cm. Therefore, the step length is 70.7 cm = 0.707 m.
Table 4.2 shows the coordinates of the body segments for both left and right halves of the
body for frame 15. The mass fractions for each segment are as follows: foot = 0.0145,
leg = 0.0465, thigh = 0.10,1/2 HAT = 0.339. The mass of HAT dominates the body COM,
but the energy changes in the lower limbs will be seen to be dominant as far as walking is
concerned (see Chapter 6).
COM analyses in three dimensions are not an easy measure to make because every seg-
ment of the body must be identified with markers and tracked with a three-dimensional
(3D) imaging system. In some studies of standing, the horizontal anterior/posterior dis-
placement of a rod attached to the pelvis has been taken as an estimate of COM movement
(Horak et al., 1992). However, in situations when a patient flexes the total body at the hip
(called a “hip strategy”) to defend against a forward fall, the pelvis moves posteriorly
considerably more than the COM (Horak and Nashner, 1986). In 3D assessments of
COM displacements, the only technique is optical tracking of markers on all segments
TABLE 4.2 Coordinates for Body Segments, Example 4.5

X (m) Y (m)
Segment Right Left Right Left
Foot 0.791 1.353 − 0.707 = 0.646 0.101 0.067
Leg 0.814 1.355 − 0.707 = 0.648 0.374 0.334
Thigh 0.787 1.402 − 0.707 = 0.695 0.708 0.691
1 2 HAT 0.721 1.424 − 0.707 = 0.717 1.124 1.122
x = 0.0145(0.791 + 0.645) + 0.0465(0.814 + 0.648) + 0.1(0.787 + 0.695) + 0.339(0.721 + 0.717) = 0.724 m
y = 0.0145(0.101 + 0.067) + 0.0465(0.374 + 0.334) + 0.1(0.708 + 0.691) + 0.339(1.124 + 1.122) = 0.937 m
ISTUDY
13 14
1. R Ankle 12. L Shoulder
9 12 2. L Ankle 13. R Ear
3. R Knee 14. L Ear
4. L Knee 15. R ASIS
21 5. R Hip 16. L ASIS
6. L Hip 17. R Iliac Crest
7. R Wrist 18. L Iliac Crest
8 19 20 11
8. R Elbow 19. R Lower Rib
9. R Shoulder 20. L Lower Rib
17 18 10. L Wrist 21. Xiphoid
11. L Elbow
15 16
10
7 5 6
Segment Mass Definition of segment
fraction COM
Head 0.081 (13 + 14)/2
Trunk 4 0.136 (9 + 12 + 21)/3
Trunk 3 0.078 ((19 + 20)/2 + 21)/2
Trunk 2 0.065 (17 + 18 + 19 + 20)/4
3 4 Trunk 1 0.078 (17 + 18 + 15 + 16)/4
Pelvis 0.142 (15 + 16)/2
Thighs 0.100 (2) 0.433 × 3 + 0.567 × 5 and
0.433 × 4 + 0.567 × 6
Legs and feet 0.060 (2) 0.606 × 1 + 0.394 × 3 and
0.606 × 2 + 0.394 × 4
1 2 Upper arms 0.028 (2) 0.436 × 8 + 0.564 × 9 and
0.436 × 11 + 0.564 × 12
Lower arms 0.022 (2) 0.682 × 7 + 0.318 × 8 and
0.682 × 10 + 0.318 × 11
Total 1.00
Figure 4.6 A 21-marker, 14-segment model to estimate the 3D center of mass of the total body in balance control
experiments. Four trunk segments were necessary to track the internal mass shifts of the thoracic/lumbar volumes.
(or as many segments as possible). MacKinnon and Winter (1993) used a seven-segment
total body estimate of the lower limbs and of the HAT to identify balance mechanisms in
the frontal plane during level walking. Jian et al. (1993) reported a 3D analysis of a similar
seven-segment estimate of the total body COM in conjunction with the center of pressure
during initiation and termination of gait and identified the motor mechanisms responsible
for that common movement.
One of the most complete measures of COM to date has been a 21-marker, 14-segment model
that has been used to determine the mechanisms of balance during quiet standing (Winter et al.,
1998). Figure 4.6 shows the location of the markers, and the accompanying table gives the def-
inition of each of the 14 segments, along with mass fraction of each segment. It is worth noting
that most of the segments are fairly rigid segments (head, pelvis, upper and lower limbs). How-
ever, the trunk is not that rigid, and it required four separate segments to achieve a reliable esti-
mate, mainly because these trunk segments undergo internal mass shifts due to respiratory and
cardiac functions. The validity of any COM estimate can be checked with the equation for the
inverted pendulum model during the movement: COP − COM = −K CÖM (Winter et al., 1998).
ISTUDY
78 ANTHROPOMETRY
COP is the center of pressure recorded from force plate data, CÖM is the horizontal acceleration
of COM in either the anterior/posterior or medial/lateral direction, and K = I/Wh, where I is the
moment of inertia of the total body about the ankles, W is body weight, and h is the height of
COM above the ankles.
4.1.7.3 Calculation of Moment of Inertia.
Example 4.6. Calculate the moment of inertia of the leg about its COM, its distal end, and its
proximal end. From Table 4.1, the mass of the leg is 0.0465 × 80 = 3.72 kg. The leg length is
given as 0.435 m. The radius of gyration/segment length is 0.302 for the COM, 0.528 for the
proximal end, and 0.643 for the distal end.
2
I 0 = 3 72 0 435 × 0 302 = 0 064 kg m2
About the proximal end,

2
I p = 3 72 0 435 × 0 528 = 0 196 kg m2
About the distal end,

2
I d = 3 72 0 435 × 0 643 = 0 291 kg m2
Note that the moment of inertia about either end could also have been calculated using the
parallel axis theorem. For example, the distance of the COM of the leg from the proximal end
is 0.433 × 0.435 = 0.188 m, and:
2
I p = I 0 + mx2 = 0 064 + 3 72 0 188 = 0 196 kg m2 0 1pc
Example 4.7. Calculate the moment of inertia of HAT about its proximal end and about its
COM. From Table 4.1, the mass of HAT is 0.678 × 80 = 54.24 kg. The HAT length is given
as 0.295 m. The radius of gyration about the proximal end/segment length is 1.456.
2
I p = 54 24 0 295 × 1 456 = 10 01 kg m2 0 1pc
From Table 4.1, the COM/segment length = 1.142 from the proximal end.
2
I 0 = I p − mx2 = 10 01 − 54 24 0 295 × 1 142 = 3 85 kg m2
We could also use the radius of gyration/segment length about the COM = 0.903.
2
I 0 = mp2 = 54 24 0 295 × 0 903 = 3 85 kg m2 0 1pc
4.2 DIRECT EXPERIMENTAL MEASURES
For more exact kinematic and kinetic calculations, it is preferable to have directly measured
anthropometric values. The equipment and techniques that have been developed have limited
capability and sometimes are not much of an improvement over the values obtained from tables.
ISTUDY
4.2 DIRECT EXPERIMENTAL MEASURES 79
(a)
x3 x4
x2
x1
s
w2 w1 w4
(b)
x3 x5
s′ w4
Figure 4.7 Balance board technique. (a) In vivo determination of mass of the location of the anatomical center of
mass of the body. (b) Mass of a distal segment. See text for details.
4.2.1 Location of the Anatomical Center of Mass of the Body

The COM of the total body, called the anatomical center of mass, is readily measured using a
balance board, as shown in Figure 4.7a. It consists of a rigid board mounted on a scale at one
end and a pivot point at the other end, or at some convenient point on the other side of the body’s
COM. There is an advantage in locating the pivot as close as possible to the COM. A more sen-
sitive scale (0–5 kg) rather than a 50- or 100-kg scale is possible, which will result in greater
accuracy. It is presumed that the weight of the balance board, w1, and its location, x1, from
the pivot are both known along with the bodyweight, w2. With the body lying prone the scale
reading is S (an upward force acting at a distance x3 from the pivot). Taking moments about
the pivot:
w1 x1 + w2 x2 = Sx3
Sx3 − w1 x1 (4.13)
x2 =
w2
4.2.2 Calculation of the Mass of a Distal Segment

The mass or weight of a distal segment can be determined by the technique demonstrated in
Figure 4.7b. The desired segment, here the leg and foot, is lifted to a vertical position so that
its COM lies over the joint center. Prior to lifting, the COM was x4 from the pivot point, with
the scale reading S. After lifting, the leg COM is x5 from the pivot, and the scale reading has
increased to S1. The decrease in the clockwise moment due to the leg movement is equal to
the increase in the scale reaction force moment about the pivot point,
ISTUDY
80 ANTHROPOMETRY
W 4 x 4 − x5 = S 1 − S x3
S1 − S x3 (4.14)
w4 =
x4 − x5
The major error in this calculation is the result of errors in x4, usually obtained from anthro-
pometric tables. To get the mass of the total limb, this experiment can be repeated with the subject
lying on his back and the limb flexed at an angle of 90 . From the mass of the total limb, we can
now subtract that of the leg and foot to get the thigh mass.
4.2.3 Moment of Inertia of a Distal Segment

The equation for the moment of inertia, described in Section 4.1.5, can be used to calculate I at a
given joint center of rotation. I is the constant of proportionality that relates the joint moment to
the segment’s angular acceleration, assuming that the proximal segment is fixed. A method called
the quick release experiment can be used to calculate I directly and requires the arrangement pic-
tured in Figure 4.8. We know that I = M/α, so if we can measure the moment M that causes an
angular acceleration α, we can calculate I directly. A horizontal force F pulls on a convenient rope
or cable at a distance y1 from the joint center and is restrained by an equal and opposite force
acting on a release mechanism. An accelerometer is attached to the leg at a distance y2 from
the joint center. The tangential acceleration a is related to the angular acceleration of the leg
α by a = y2α.
With the forces in balance as shown, the leg is held in a neutral position and no acceleration
occurs. If the release mechanism is actuated, the restraining force suddenly drops to zero and the
net moment acting on the leg is Fy1, which causes an instantaneous acceleration α. F and a can be
recorded on a dual-beam storage oscilloscope; most pen recorders have too low a frequency
response to capture the acceleration impulse. The moment of inertia can now be calculated,
M Fy y
I= = 1 2 (4.15)
α a
Figure 4.8 shows the sudden burst of acceleration accompanied by a rapid decrease in the
applied force F. This force drops after the peak of acceleration and does so because the forward
displacement of the limb causes the tension to drop in the pulling cable. A convenient release
mechanism can be achieved by suddenly cutting the cable or rope that holds back the leg.
Force
F
transducer
a
y1 y2
Dual–beam
F storage scope
a
Accelerom.
Amplif.
Release
mechanism Amplif.
Figure 4.8 Quick-release technique for the determination of the mass moment of inertia of a distal segment. Force
F applied horizontally results, after release of the segment, in an initial acceleration a. Moment of inertia can then be
calculated from F, y1, y2.
ISTUDY
4.2 DIRECT EXPERIMENTAL MEASURES 81
The sudden accelerometer burst can also be used to trigger the oscilloscope sweep so that the
rapidly changing force and acceleration can be captured.
More sophisticated experiments have been devised to measure more than one parameter simul-
taneously. Such techniques were developed by Hatze (1975) and are capable of determining the
moment of inertia, the location of the COM, and the damping coefficient simultaneously.
4.2.4 Joint Axes of Rotation

Markers attached to the body are usually placed to represent our best estimate of a joint center.
However, because of anatomical constraints, our location can be somewhat in error. The lateral
malleolus, for example, is a common location for ankle joint markers. However, the articulation
of the tibial/talus surfaces is such that the distal end of the tibia (and the fibula) move in a small arc
over the talus. The true axis of rotation is actually a few centimeters distal of the lateral malleolus.
Even more drastic differences are evident at some other joints. The hip joint is often identified in
the sagittal plane by a marker on the upper border of the greater trochanter. However, it is quite
evident that the marker is somewhat more lateral than the center of the hip joint such that internal
and external rotations of the thigh relative to the pelvis may cause considerable errors, as will
abduction/adduction at that joint.
Thus, it is important that the true axes of rotation be identified relative to anatomical markers
that we have placed on the skin. There are two primary methods to estimate joint centers, pre-
dictive and functional models. In a predictive model, a regression equation derived from medical
imaging is used to estimate the location of the internal joint center based on the location of one or
more surface markers on the segment. For example, Harrington et al. (2007) used created regres-
sion equations based on the subject characteristics of pelvic width, pelvic depth, and leg length to
estimate the 3D locations of the hip joint center using markers placed on anatomical landmarks.
In the functional model, joint center is estimated based on the movement of markers on adjacent
segments. There are a variety of functional models, and the specifics of these techniques are in
part dependent on the degrees of freedom of the joint of interest. Geometric sphere fitting is one
such functional method that is commonly applied to determine the center of the hip joint (Leardini
et al., 1999). In this method, the subject has markers attached to the pelvis and along the thigh.
The subject moves the limb through sequential flexion and extension, abduction and adduction,
and circumduction to create a spherical path of the thigh markers. The path of the markers is
tracked and the center of this motion is calculated and defined as the hip joint center. Geometric
sphere fitting algorithms may be best applied to the ball and socket motion of the hip joint,
whereas a more constrained approach may better fit the knee or ankle joint.
One example of this functional technique is described here. Figure 4.9 shows two segments in
a planar movement. First, they must be translated and rotated in space so that one segment is fixed
in space and the second rotates as shown. At any given instant in time, the true axis of rotation is at
(xc, yc) within the fixed segment, and we are interested in the location of (xc, yc) relative to ana-
tomical coordinates (x3, y3) and (x4, y4) of that segment. Markers (x1, y1) and (x2, y2) are located as
shown; (x1, y1) has an instantaneous tangential velocity V and is located at a radius R from the axis
of rotation. From the line joining (x1, y1) to (x2, y2), we calculate the angular velocity of the rotat-
ing segment ωz . With one segment fixed in space, ωz is nothing more than the joint angular
velocity,
V = ωz × R (4.16a)
or, in Cartesian coordinates,
V x i + V y j = Ry ωz i − Rx ωz j
ISTUDY
82 ANTHROPOMETRY
(x1, y1)
Vx
Vy V
Ry
ωz
R
(x2, y2)
xc, yc Rx
fixed
(x3, y3) segment (x4, y4)
Figure 4.9 Technique used to calculate the axis of rotation between two adjacent segments xc, yc. Each segment
must have two markers in the plane of movement. After data collection, the segments are rotated and translated so
that one segment is fixed in space. Thus, the moving-segment kinematics reflects the relative movement between the
two segments, and the axis of rotation can be located relative to the anatomical location of markers on that fixed
segment. See text for complete details.
Therefore,
V x = Ry ωz and V y = − Rx ωz (4.16b)
Since Vx, Vy, and ωz can be calculated from the marker trajectory data, Ry and Rx can be deter-
mined. Since x1, y1 is known, the axis of rotation xc, yc can be calculated. Care must be taken when
ωz approaches 0 or reverses its polarity, because R, as calculated by Equation (4.16a), becomes
indeterminate or falsely approaches very large values. In practice, we have found errors become
significant when ωz falls below 0.5 r/s.
Currently, the International Society of Biomechanics recommends the use of functional
algorithms for individuals with sufficient range of motion to perform the necessary calibration
movements. For individuals without adequate motion, predictive models are recommended
(Wu et al., 2002).
4.3 MUSCLE ANTHROPOMETRY
Before we can calculate the forces produced by individual muscles during normal movement, we
usually need some dimensions from the muscles themselves. If muscles of the same group share
the load, they probably do so proportionally to their relative cross-sectional areas. Also, the
mechanical advantage of each muscle can be different, depending on the moment arm length
at its origin and insertion, and on other structures beneath the muscle or tendon that alter the angle
of pull of the tendon.
4.3.1 Cross-Sectional Area of Muscles

The functional or physiologic cross-sectional area (PCA) of a muscle is a measure of the number
of sarcomeres in parallel with the angle of pull of the muscles. In pennate muscles, the fibers act at
an angle from the long axis and, therefore, are not as
ISTUDY
4.3 MUSCLE ANTHROPOMETRY 83
muscle. The angle between the long axis of the muscle and the fiber angle is called pennation
angle. In parallel-fibered muscle, the PCA is:
m 2
PCA = cm (4.17)
dl
where
m = mass of muscle fibers, g
d = density of muscle, g/cm3, =1.056 g/cm3
l = length of muscle fibers, cm.
In pennate muscles, the physiological cross-sectional area becomes:
m cos θ 2
PCA = cm (4.18)
dl
where θ is the pennation angle, which increases as the muscle shortens.
Wickiewcz et al. (1983), using data from three cadavers, measured muscle mass, fiber
lengths, and pennation angle for 27 muscles of the lower extremity. Representative values
are given in Table 4.3. The PCA as a percentage of the total cross-sectional area of all mus-
cles crossing a given joint is presented in Table 4.4. In this way, the relative potential con-
tribution of a group of agonist muscles can be determined, assuming that each is generating
the same stress. Note that a double-joint muscle, such as the gastrocnemius, may represent
different percentages at different joints because of the different total PCA of all muscles
crossing each joint.
4.3.2 Change in Muscle Length During Movement

A few studies have investigated the changes in the length of muscles as a function of the angles
of the joints they cross. Grieve et al. (1978), in a study on eight cadavers, reported percentage
length changes of the gastrocnemius muscle as a function of the knee and ankle angle. The
resting length of the gastrocs was assumed to be when the knee was flexed 90 and the ankle
was in an intermediate position, neither plantarflexed nor dorsiflexed. With 40 plantarflexion,
the muscle shortened 8.5% and linearly changed its length to a 4% increase at 20 dorsiflexion.
An almost linear curve described the changes at the knee: 6.5% at full extension to a 3%
decrease at 150 flexion.
TABLE 4.3 Mass, Length, and PCA of Some Muscles

Muscle Mass (g) Fiber Length (cm) PCA (cm2) Pennation Angle ( )
Sartorius 75 38 1.9 0
Biceps femoris (long) 150 9 15.8 0
Semitendinosus 75 16 4.4 0
Soleus 215 3.0 58 30
Gastrocnemius 158 4.8 30 15
Tibialis posterior 55 2.4 21 15
Tibialis anterior 70 7.3 9.1 5
Rectus femoris 90 6.8 12.5 5
Vastus lateralis 210 6.7 30 5
Vastus medialis 200 7.2 26 5
Vastus intermedius 180 6.8 25 5
ISTUDY
84 ANTHROPOMETRY
TABLE 4.4 Four Percent PCA of Muscles Crossing Ankle, Knee, and Hip Joints
Ankle Knee Hip
Muscle %PCA Muscle %PCA Muscle %PCA
Soleus 41 Gastrocnemius 19 Iliopsoas 9
Gastrocnemius 22 Biceps femoris (small) 3 Sartorius 1
Flexor hallucis longus 6 Biceps femoris (long) 7 Pectineus 1
Flexor digitorum longus 3 Semitendinosus 3 Rectus femoris 7
Tibialis posterior 10 Semimembranosus 10 Gluteus maximus 16
Peroneus brevis 9 Vastus lateralis 20 Gluteus medius 12
Tibialis anterior 5 Vastus medialis 15 Gluteus minimus 6
Extensor digitorum longus 3 Vastus intermedius 13 Adductor magnus 11
Extensor hallucis longus 1 Rectus femoris 8 Adductor longus 3
Sartorius 1 Adductor brevis 3
Gracilis 1 Tensor fasciae latae 1
Biceps femoris (long) 6
Semitendinosus 3
Semimembranosus 8
Piriformis 2
Lateral rotators 13
4.3.3 Force per Unit Cross-Sectional Area (Stress)

A wide range of stress values for skeletal muscles has been reported (Haxton, 1944; Alexander
and Vernon, 1975; Maughan et al., 1983). Most of these stress values were measured during iso-
metric conditions and range from 20 to 100 N/cm2. These higher values were recorded in pennate
muscles, which are those whose fibers lie at an angle from the main axis of the muscle. Such an
orientation effectively increases the cross-sectional area above that measured and used in the
stress calculation. Haxton (1944) related force to stress in two pennate muscles (gastrocs and
soleus) and found stresses as high as 38 N/cm2. Dynamic stresses have been calculated in the
quadriceps during running and jumping to be about 70 N/cm2 (based on a peak knee extensor
moment of 210 Nm in adult males) and about 100 N/cm2 in isometric maximum voluntary con-
tractions (MVCs) (Maughan et al., 1983).
4.3.4 Mechanical Advantage of Muscle

The origin and insertion of each muscle define the angle of pull of the tendon on the bone and,
therefore, the mechanical leverage it has at the joint center. Each muscle has its unique moment
arm length, which is the length of a line normal to the muscle passing through the joint center.
This moment arm length changes with the joint angle. One of the first studies done in this area
(Smidt, 1973) reported the average moment arm length (26 subjects) for the knee extensors and
for the hamstrings acting at the knee. Both these muscle groups showed an increase in the
moment length as the knee was flexed, reaching a peak at 45 , then decreasing again as flexion
increased to 90 . As musculoskeletal imaging has improved, researchers are able to measure
in vivo moment arms during dynamic activities. This is an important consideration as muscle
force during activity may affect the distance between its line of action and center of joint rotation.
Accurate estimation of moment arm length throughout the course of range of motion and during
active muscle contraction is a key component for simulation studies, as discussed in detail in
Chapter 10.
ISTUDY
4.3 MUSCLE ANTHROPOMETRY 85
4.3.5 Multijoint Muscles

A large number of the muscles in the human body pass over more than one joint. In the lower
limbs, the hamstrings are extensors of the hip and flexors of the knee, the rectus femoris is a com-
bined hip flexor and knee extensor, and the gastrocnemius is knee flexors and ankle plantarflex-
ors. The fiber length of many of these muscles may be insufficient to allow a complete range of
movement of both joints involved. Elftman (1966) has suggested that many normal movements
require lengthening at one joint simultaneously with shortening at the other. Consider the action
of the rectus femoris, for example, during early swing in running. This muscle shortens as a result
of hip flexion and lengthens at the knee as the leg swings backward in preparation for swing. The
tension in the rectus femoris simultaneously creates a flexor hip moment (positive work) and an
extensor knee moment to decelerate the swinging leg (negative work) and start accelerating it
forward. In this way, the net change in muscle length is reduced compared with 2 equivalent
single-joint muscles, and excessive positive and negative work within the muscle can be reduced.
A double-joint muscle could even be totally isometric in such situations and would effectively be
transferring energy from the leg to the pelvis in the example just described. In running during the
critical push-off phase, when the plantarflexors are generating energy at a high rate, the knee is
continuing to extend. Thus, the gastrocnemii may be essentially isometric (they may appear to be
shortening at the distal end and lengthening at the proximal end). Similarly, toward the end of
swing in running, the knee is rapidly extending while the hip has reached full flexion and is begin-
ning to reverse (i.e. it has an extensor velocity). Thus, the hamstrings appear to be rapidly length-
ening at the distal end and shortening at the proximal end, with the net result that they may be
lengthening at a slower rate than a single joint would.
It is also critical to understand the role of the major biarticulate muscles of the lower limb dur-
ing stance phase of walking or running. Figure 4.10 shows the gastrocnemii, hamstrings, and
rectus femoris and their moment-arm lengths at their respective proximal and distal ends. The
hamstrings have a 5-cm moment-arm at the ankle and 3.5-cm moment-arm at the knee. Thus,
6 – 7 cm 5 cm
3 – 5 cm
3 – 5 cm
4 cm
5 cm
Figure 4.10 Three major biarticulate muscles of the lower limb. Shown are the gastrocnemii, hamstrings, and
rectus femoris, and their moment-arm lengths at their proximal and distal ends. These moment-arms are critical
to the functional role of these muscle groups during weight bearing;
ISTUDY
86 ANTHROPOMETRY
when they are active during stance, their contribution to the ankle extensor moment is about 50%
greater than their contribution to the knee flexor moment. The net effect of these two contribu-
tions is to cause the leg to rotate posteriorly and prevent the knee from collapsing. The
hamstrings, with the exception of the short head of the biceps femoris, have moment-arms of
6–7 cm at the hip but only 3.5 cm at the knee. Thus, when these muscles are active during stance,
their contribution to hip extension is about twice their contribution to knee flexion. The net effect
of these two actions is to cause the thigh to rotate posteriorly and prevent the knee from collap-
sing. Finally, the rectus femoris is the only biarticulate muscle of the large quadriceps group, and
its moment-arm at the hip is slightly larger than at the knee. However, the quadriceps activate as a
group, and because the uniarticulate quadriceps comprise 84% of the PCA of the quadriceps (see
Table 4.3), the dominant action is knee extension. Thus, the net effect of the major biarticulate
muscles of the lower limb is extension at all three joints, and therefore they contribute, along with
all the uniarticulate extensors, to defending against a gravity-induced collapse. The algebraic
summation of all three moments during stance phase of gait has been calculated and has been
found to be dominantly extensor (Winter, 1984). This summation has been labeled the support
moment and is discussed further in Section 11.1.
4.4 PROBLEMS BASED ON ANTHROPOMETRIC DATA
1. (a) Calculate the average body density of a young adult whose height is 1.68 m and whose
mass is 68.5 kg. Answer: 1.059 kg/L.
(b) For the adult in (a), determine the density of the forearm and use it to estimate the mass
of the forearm that measures 24.0 cm from the ulnar styloid to the elbow axis. Circum-
ference measures (in cm) taken at 1-cm intervals starting at the wrist are 20.1, 20.3,
20.5, 20.7, 20.9, 21.2, 21.5, 21.9, 22.5, 23.2, 23.9, 24.6, 25.1, 25.7, 26.4, 27.0,
27.5, 27.9, 28.2, 28.4, 28.4, 28.3, 28.2, and 28.0. Assuming the forearm to have a cir-
cular cross-sectional area over its entire length, calculate the volume of the forearm and
its mass. Compare the mass as calculated with that estimated using averaged anthro-
pometric data (Table 4.1). Answer: Forearm density = 1.13 kg/L; volume = 1.174 L;
mass = 1.33 kg. Mass calculated from Table 4.1 = 1.10 kg.
(c) Calculate the location of the COM of the forearm along its long axis and give its dis-
tance from the elbow axis. Compare that with the COM as determined from Table 4.1.
Answer: COM = 10.34 cm from the elbow; from Table 4.1, COM = 10.32 cm from
the elbow.
(d) Calculate the moment of inertia of the forearm about the elbow axis. Then calculate its
radius of gyration about the elbow and compare it with the value calculated from
Table 4.1. Answer: Ip = 0.0201 kg m2; radius of gyration = 12.27 cm; radius of gyra-
tion from Table 4.1 = 12.62 cm.
2. (a) From the data listed in Table A.3b in Appendix A, calculate the COM of the lower limb
for frame 70. Answer: x = 1.755 m; y = 0.522 m.
(b) Using your data of stride length (Problem 2(f ) in Section 3.8) and a stride time of
68 frames, create an estimate (assuming symmetrical gait) of the coordinates of
the left half of the body for frame 30. From the segment centers of mass
(Table A.4), calculate the COM of the right half of the body (foot + leg + high + 1/
2 HAT), and of the left half of the body (using segment data suitably shifted in time
and space). Average the two centers of mass to get the COM of the total body for
frame 30. Answer: x = 1.025 m, y = 0.904 m.
3. (a) Calculate the moment of inertia of the HAT about its COM for the subject described in
Appendix A. Answer: I0 = 1.96 kg m2.
ISTUDY
4.5 REFERENCES 87
(b) Assuming that the subject is standing erect with the two feet together, calculate the
moment of inertia of HAT about the hip joint, the knee joint, and the ankle joint.
What does the relative size of these moments of inertia tell us about the relative
magnitude of the joint moments required to control the inertial load of HAT.
Answer: Ih = 5.09 kg m2, Ik = 15.78 kg m2, Ia = 42.31 kg m2.
(c) Assuming that the COM of the head is 1.65 m from the ankle, what percentage does it
contribute to the moment of inertia of HAT about the ankle? Answer: Ihead about
ankle =12.50 kg m2, which is 29.6% of Ihat about the ankle.
4. (a) Calculate the moment of inertia of the lower limb of the subject in Appendix A about
the hip joint. Assume that the knee is not flexed and the foot is a point mass located
6 cm distal to the ankle. Answer: Ilower limb about hip =1.39 kg m2.
(b) Calculate the increase in the moment of inertia as calculated in (a) when a ski boot is
worn. The mass of the ski boot is 1.8 kg, and assume it to be a point mass located 1 cm
distal to the ankle. Answer: Iboot about hip = 1.01 kg m2.
4.5 REFERENCES
Alexander, R. M. and A. Vernon. “The Dimensions of Knee and Ankle Muscles and the Forces They Exert,”
J. Hum. Mov. Stud. 1: 115–123, 1975.
Contini, R. “Body Segment Parameters, Part II,” Artif. Limbs 16: 1–19, 1972.
Dempster, W. T. “Space Requirements of the Seated Operator,” WADC-TR-55-159, Wright Patterson Air
Force Base, 1955.
Dempster, W. T., W. C. Gabel, and W. J. L. Felts. “The Anthropometry of Manual Work Space for the Seated
Subjects,” Am. J. Phys. Anthrop. 17: 289–317, 1959.
Drillis, R. and R. Contini. “Body Segment Parameters,” Rep. 1163-03, Office of Vocational Rehabilitation,
Department of Health, Education, and Welfare, New York, 1966.
Dumas, R. and J. Wojtusch. “Estimation of the Body Segment Inertial Parameters for the Rigid Body Bio-
mechanical Models Used in Motion Analysis,” in Handbook of Human Motion, B. Müller and S. Wolf,
Eds. (Springer, Cham, 2018).
Dumas, R., L. Chèze, and J. P. Verriest. “Adjustments to McConville et al. and Young et al. Body Segment
Inertial Parameters,” J. Biomech. 40(3): 543–553, 2007.
Elftman, H. “Biomechanics of Muscle, with Particular Application to Studies of Gait,” J. Bone Joint Surg.
48-A: 363–377, 1966.
Grieve, D. W., P. R. Cavanagh, and S. Pheasant. “Prediction of Gastrocnemius Length from Knee and Ankle
Joint Posture,” in Biomechanics, Vol. VI-A, E. Asmussen and K. Jorgensen, Eds. (University Park Press,
Baltimore, MD, 1978), pp. 405–412.
Harrington, M. E., A. B. Zavatsky, S. E. Lawson, Z. Yuan, and T. N. Theologis. “Prediction of the Hip Joint
Centre in Adults, Children, and Patients with Cerebral Palsy Based on Magnetic Resonance Imaging,”
J. Biomech. 40(3): 595–602, 2007.
Hatze, H. “A New Method for the Simultaneous Measurement of the Moment of Inertia, the Damping Coef-
ficient and the Location of the Center of Mass of a Body Segment In Situ,” Eur. J. Appl. Physiol. 34: 217–
266, 1975.
Haxton, H. A. “Absolute Muscle Force in the Ankle Flexors of Man,” J. Physiol. 103: 267–273, 1944.
Horak, F. B. and L. M. Nashner. “Central Programming of Postural Movements: Adaptation to Altered Sup-
port Surface Configurations,” J. Neurophysiol. 55: 1369–1381, 1986.
Horak, F. B., J. G. Nutt, and L. M. Nashner. “Postural Inflexibility in Parkinsonian Subjects,” J. Neurol. Sci.
111: 46–58, 1992.
Jian, Y., D. A. Winter, M. G. Ishac, and L. Gilchrist. “Trajectory of the Body COG and COP During Initiation
and Termination of Gait,” Gait Posture 1: 9–22, 1993.
ISTUDY
88 ANTHROPOMETRY
Leardini, A., A. Cappozzo, F. Catani et al. “Validation of a Functional Method for the Estimation of Hip Joint
Centre Location,” J. Biomech. 32(1): 99–103, 1999.
de Leva, P. “Adjustments to Zatsiorsky-Seluyanov’s Segment Inertia Parameters,” J. Biomech. 29(9): 1223–
1230, 1996.
MacKinnon, C. D. and D. A. Winter. “Control of Whole Body Balance and Posture in the Frontal Plane Dur-
ing Walking,” J. Biomech. 26: 633–644, 1993.
Maughan, R. J., J. S. Watson, and J. Weir. “Strength and Cross-Sectional Area of Human Skeletal Muscle,”
J. Physiol. 338: 37–49, 1983.
Smidt, G. L. “Biomechanical Analysis of Knee Flexion and Extension,” J. Biomech. 6: 79–92, 1973.
Wickiewcz, T. L., R. R. Roy, P. L. Powell, and V. R., Edgerton. “Muscle Architecture of the Human Lower
Limb,” Clin. Orthop. Rel. Res. 179: 275–283, 1983.
Winter, D. A. “Kinematic and Kinetic Patterns in Human Gait: Variability and Compensating Effects,” Hum.
Mov. Sci. 3: 51–76, 1984.
Winter, D. A., A. E. Patla, F. Prince, M. G. Ishac, and K. Gielo-Perczak. “Stiffness Control of Balance in Quiet
Standing,” J. Neurophysiol. 80: 1211–1221, 1998.
Wu, G., S. Siegler, P. Allard et al. “ISB Recommendation on Definitions of Joint Coordinate System of Var-
ious Joints for the Reporting of Human Joint Motion--Part I: Ankle, Hip, and Spine. International Society
of Biomechanics,” J. Biomech. 35(4): 543–548, 2002.
ISTUDY
5
KINETICS: FORCES AND MOMENTS
OF FORCE
1
2
3
5.0 BIOMECHANICAL MODELS
Chapter 3 dealt at length with the movement itself, without regard to the forces that cause the
movement. The study of these forces and the resultant energetics is called kinetics. Knowledge
of the patterns of the forces is necessary for an understanding of the cause of any movement.
Transducers have been developed that can be implanted surgically to measure the force exerted by
a muscle at the tendon. However, such techniques have applications only in animal experiments
and, even then, only to a limited extent. It, therefore, remains that we attempt to calculate these
forces indirectly, using readily available kinematic and anthropometric data. The process by
which the reaction forces and muscle moments are calculated is called link-segment modeling.
Such a process is depicted in Figure 5.1. If we have a full kinematic description, accurate anthro-
pometric measures, and the external forces, we can calculate the joint reaction forces and muscle
moments. This prediction is called an inverse solution and is a very powerful tool in gaining
insight into the net summation of all muscle activity at each joint. Such information is very useful
to the surgeon, therapist, athletic trainer, kinesiologist, exercise scientist, and coach in their diag-
nostic assessments. The effect of training, therapy, or surgery is extremely evident at this level of
assessment, although it is often obscured in the original kinematics.
5.0.1 Link-Segment Model Development

Since we are not able to measure the joint forces and moments of force directly, a model is
required to calculate them indirectly. The validity of any assessment is only as good as the model
itself. Accurate measures of segment masses, centers of mass (COMs), joint centers, and
moments of inertia are required. Such data can be obtained from statistical tables based on
the person’s height, weight, and, sometimes, sex, as was detailed in Chapter 4. A limited number
of these variables can be measured directly, but some of the techniques are time-consuming and
†
Deceased.

Companion website:
ISTUDY
90 KINETICS: FORCES AND MOMENTS OF FORCE
Kinematic Joint reaction

ak
data Mham Forces
θa
θk
Kinetic Net muscle

data moments
L
Mquads
θsh
Anthropometric Kinetic and
data Msh potential energy
g Mmuscle = Mquads – Mham
Figure 5.1 Schematic diagram of the relationship between kinematic, kinetic, and anthropometric data and the
calculated forces, moments, energies, and powers using an inverse solution of a link-segment model.
have limited accuracy. Regardless of the source of the anthropometric data, the following
assumptions are made with respect to the model:
1. Each segment has a fixed mass located as a point mass at its COM (which will be the center
of gravity in the vertical direction).
2. The location of each segment’s COM remains fixed during the movement.
3. The joints are considered to be hinge (or ball-and-socket) joints.
4. The mass moment of inertia of each segment about its mass center (or about either proximal
or distal joints) is constant during the movement.
5. The length of each segment remains constant during the movement (e.g. the distance
between hinge or ball-and-socket joints remains constant).
Figure 5.2 shows the equivalence between the anatomical and the link-segment models for the
lower limb. The segment masses m1, m2, and m3 are considered to be concentrated at points. The
distance from the proximal joint to the mass centers is considered to be fixed, as are the length of
the segments and each segment’s moment of inertia I1, I2, and I3 about each COM.
5.0.2 Forces Acting on the Link-Segment Model

1. Gravitational Forces. The forces of gravity act downward through the COMs of each seg-
ment and are equal to the magnitude of the mass times acceleration due to gravity (normally
9.8 m/s2).
2. Ground Reaction or External Forces. Any external forces must be measured by a force
transducer. Such forces are distributed over an area of the body (such as the ground reac-
tion forces under the area of the foot). In order to represent such forces as vectors, they
must be considered to act at a point that is usually called the center of pressure (COP).
A suitably constructed force plate, for example, yields signals from which the COP can be
calculated.
ISTUDY
5.0 BIOMECHANICAL MODELS 91
1 1
I1 m1
2 2
I2 m2
3 3
m3
I3
Anatomical Link-segment
model model
Figure 5.2 Relationship between anatomical and link-segment models. Joints are replaced by hinge (pin) joints
and segments are replaced by masses and moments of inertia located at each segment’s center of mass.
3. Muscle and Ligament Forces. The net effect of muscle activity at a joint can be calculated
in terms of net muscle moments. If a co-contraction is taking place at a given joint, the
analysis yields only the net effect of both agonist and antagonistic muscles. Also, any fric-
tion effects at the joints or within the muscle cannot be separated from this net value.
Increased friction merely reduces the effective “muscle” moment; the muscle contractile
elements themselves are actually creating moments higher than that analyzed at the tendon.
However, the error at low and moderate speeds of movement is usually only a few percent.
At the extreme range of movement of any joint, passive structures such as joint capsule and
ligaments come into play to contain the range. The moments generated by these tissues will
add to or subtract from those generated by the muscles. Thus, unless the muscle is silent, it
is impossible to determine the contribution of these passive structures.
5.0.3 Joint Reaction Forces and Bone-on-Bone Forces

The three forces described in the preceding sections constitute all the forces acting on the total
body system itself. However, our analysis examines the segments one at a time and, therefore,
must calculate the reaction between segments that occur within the joints (Paul, 1966). A free-
body diagram of each segment is required, as shown in Figure 5.3. Here the original link-segment
model is broken into its segmental parts. For convenience, we make the break at the joints and the
forces that act across each joint must be shown on the resultant free-body diagram. This proce-
dure now permits us to look at each segment and calculate all unknown joint reaction forces. In
accordance with Newton’s 3rd law, there is an equal and opposite force acting at each hinge joint
in our model. For example, when a leg is held off the ground in a static condition, the foot is
exerting a downward force on the joint capsule, tendons, and ligaments crossing the ankle joint.
This is seen as a downward force acting on the leg equal to the weight of the foot. Likewise, those
passive joint tissues are exerting an equal upward force on
ISTUDY
Ry1
Rx1
1
1 M1
I1 m1
I1 m1
Rx2
2 M2
Ry2 Ry2
2
2 Rx2
I2 m2 I2 m2
M3
Rx3 Ry3
3 3 Rx3
m3 3
I3 Ry3 m3
I3
Link - segment Free - body

model diagram
Figure 5.3 Relationship between the free-body diagram and the link-segment model. Each segment is “broken” at
the joints, and the reaction forces and moments of force acting at each joint are indicated.
Considerable confusion exists regarding the relationship between joint reaction forces and
joint bone-on-bone forces. The latter forces are the actual forces seen across the articulating
surfaces and include the effect of muscle activity and the action of joint capsules and ligaments.
Actively contracting muscles pull the articulating surfaces together, creating compressive
forces and, sometimes, shear forces. In addition, when passive joint structures reach their max-
imal length they can also create compressive and shear forces. In the simplest situation, the
bone-on-bone force equals the active force due to the muscles and passive structures plus
the joint reaction forces. Figure 5.4 illustrates these differences. Case 1 has the lower segment
with a weight of 100 N hanging passively from the muscles originating in the upper segment.
The two muscles are not contracting but are, assisted by joint capsule and ligamentous tissue,
pulling upward with an equal and opposite 100 N. The link-segment model shows these equal
and opposite reaction forces. The bone-on-bone force is zero, indicating that the joint articulat-
ing surfaces are under neither tension nor compression. In case 2, there is an active contraction
of the muscles, so that the total upward force is now 170 N. The bone-on-bone force is 70 N
compression. This means that a force of 70 N exists across the articulating surfaces. As far
as the lower segment is concerned, there is still a net reaction force of 100 N upward (170 N
upward through muscles and 70 N downward across the articulating surfaces). The lower seg-
ment is still acting downward with a force of 100 N; thus, the free-body diagram remains the
same. Generally, the anatomy is not as simple as depicted. More than one muscle is usually
active on each side of the joint, so it is difficult to apportion the forces among the muscles. Also,
the exact angle of pull of each tendon, joint capsule, and ligament in addition to the geometry of
the articulating surfaces is not always readily available and change throughout the joints range
of motion. Thus, a much more complex free-body diagram must be used, and the techniques and
examples are described in Section 5.3.
ISTUDY
5.1 BASIC LINK-SEGMENT EQUATIONS – THE FREE-BODY DIAGRAM 93
Case 1 Case 2
m1
85
85
50 50
70
100
100
100 m2 100
Bone–on–bone Bone–on–bone
Force = 0 Force = 70 (compress.)
Figure 5.4 Diagrams to illustrate the difference between joint reaction forces and bone-on-bone forces. In both
cases, the reaction force is 100 N acting upward on M2 and downward on M1. With no muscle activity (case 1), the
bone-on-bone force is zero; with muscle activity (case 2) it is 70 N.
5.1 BASIC LINK-SEGMENT EQUATIONS – THE FREE-BODY DIAGRAM
Each body segment acts independently under the influence of reaction forces and muscle
moments, which act at either end, plus the forces due to gravity (Bresler and Frankel, 1950). Con-
sider the planar movement of a segment in which the kinematics, anthropometrics, and reaction
forces at the distal end are known (see Figure 5.5).
Known
ax, ay = acceleration of segment COM
θ = angle of segment in plane of movement
α = angular acceleration of segment in plane of movement
Ryp
Rxp
Mp
2
d
∝
ay
a
ax
m
1
d
θ
mg
Rxd
Md
Ryd
Figure 5.5 Complete free-body diagram of a single segment, showing reaction and gravitational forces, net
moments of force, and all linear and angular accelerations.
ISTUDY
Rxd, Ryd = reaction forces acting at distal end of segment, usually determined from a prior anal-
ysis of the proximal forces acting on distal segment
Md = net muscle moment acting at distal joint, usually determined from an analysis of the
proximal muscle acting on distal segment
Unknown
Rxp, Ryp = reaction forces acting at proximal joint
Mp = net muscle moment acting on segment at proximal joint
Equations
1. ΣFx = max
Rxp − Rxd = max (5.1)
2. ΣFy = may
Ryp − Ryd − mg = may (5.2)
3 About the segment COM, ΣM = I 0 α (5.3)

Note that the muscle moment at the proximal end cannot be calculated until the proximal reac-
tion forces Rxp and Ryp have first been calculated.
Example 5.1 (see Figure 5.6). In a static situation, a person is standing on one foot on a force
plate. The ground reaction force is found to act 4 cm anterior to the ankle joint. Note that con-
vention has the ground reaction force Ry1 always acting upward. We also show the horizontal
reaction force Rx1 to be acting in the positive direction (to the right). If this force actually acts
to the left, it will be recorded as a negative number. The subject’s mass is 60 kg, and the mass
of the foot is 0.9 kg. Calculate the joint reaction forces and net muscle moment at the ankle.
Ry1 = body weight = 60 × 9.8 = 588 N.
1. ΣFx = max,
Rx2 + Rx1 = max = 0
Note that this is a redundant calculation in static conditions.
Ry2
Rx2
M2
6 6
Rx1
4 4 0.9 g
Ry1 Ry1
Figure 5.6 Anatomical and free-body diagram of foot during weight bearing.
ISTUDY
2. ΣFy = may,
Ry2 + Ry1 − mg = may
Ry2 + 588 − 0 9 × 9 8 = 0
Ry2 = − 579 2 N
The negative sign means that the force acting on the foot at the ankle joint acts downward. This
is not surprising because the entire body weight, less that of the foot, must be acting downward on
the ankle joint.
3. About the COM, ΣM = I0α,
M 2 − Ry1 × 0 02 − Ry2 × 0 06 = 0
M 2 = 588 × 0 02 + − 579 2 × 0 06 = − 22 99 N m
The negative sign means that the real direction of the muscle moment acting on the foot at the
ankle joint is clockwise, which means that the plantarflexors are active at the ankle joint to
maintain the static position. These muscles have created an action force that resulted in the
ground reaction force that was measured, and whose COP was 4 cm anterior to the
ankle joint.
Example 5.2 (see Figure 5.7). From the data collected during the swing of the foot, calculate
the muscle moment and reaction forces at the ankle. The subject’s mass was 80 kg and the ankle-
metatarsal length was 20.0 cm. From Table 4.1, the inertial characteristics of the foot are
calculated:
Ry1
Rx1
M1
9.85 –6.62 m/s2
9.07 m/s2
21.69 rad/s2
9.85 1.16 g
1.95
Figure 5.7 Free-body diagram of foot during swing showing the linear accelerations of the center of mass and the
angular acceleration of the segment. Distances are in centimeters. Three unknowns, Rx1, Ry1, and M1, are to be
calculated assuming a positive direction as shown.
ISTUDY
m = 0 0145 × 80 = 1 16 kg
ρ0 = 0 475 × 0 20 = 0 095 m
I 0 = 1 16 0 095 2 = 0 0105 kg m2
α = 21 69 rad s2
1. ΣFx = max,
Rx1 = 1 16 × 9 07 = 10 52 N
2. ΣFy = may,
Ry1 − 1 16g = m − 6 62
Ry1 = 1 16 × 9 8 − 1 16 × 6 62 = 3 69 N
3. At the COM of the foot, ΣM = I0α,
M 1 − Rx1 × 0 0985 − Ry1 × 0 0195 = 0 0105 × 21 69
M1 = 0 0105 × 21 69 + 10 52 × 0 0985 + 3 69 × 0 0195

= 0 23 + 1 04 + 0 07 = 1 34 N m
Discussion
1. The horizontal reaction force of 10.52 N at the ankle is the cause of the horizontal accel-
eration that we calculated for the foot.
2. The foot is decelerating its upward rise at the end of lift-off. Thus, the vertical reaction force
at the ankle is somewhat less than the static gravitational force.
3. The ankle muscle moment is positive, indicating net dorsiflexor activity (tibialis anterior),
and most of this moment (1.04 out of 1.34 N m) is required to cause the horizontal accel-
eration of the foot’s center of gravity, with very little needed (0.23 N m) to angularly
accelerate the low moment of inertia of the foot.
Example 5.3 (see Figure 5.8). For the same instant in time, calculate the muscle moments and
reaction forces at the knee joint. The leg segment was 43.5 cm long.
m = 0 0465 × 80 = 3 72 kg
ρ0 = 0 302 × 0 435 = 0 131 m
2
I 0 = 3 72 0 131 = 0 0638 kg m2
α = 36 9 rad s2
From Example 5.2, Rx1 = 10.52 N, Ry1 = 3.69 N, and M1 = 1.34N m
ISTUDY
Ry2
Rx2
M2
–4.21 m/s2
12·9
–0.3 m/s2
18·5 14·2
3·72 g
16·9 36·9 rad/s2
Rx1
M1
Ry1
Figure 5.8 Free-body diagram of leg at the same instant in time as the foot in Figure 5.7. Linear and angular
accelerations are as shown. Distances are in centimeters. The distal end and reaction forces and moments have
been reversed, recognizing Newton’s third law. Again, the three unknowns Rx2, Ry2, and M2 are assumed to be
positive with direction as indicated.
1. ΣFx = max,
Rx2 − Rx1 = max
Rx2 = 10 52 + 3 72 − 0 03 = 10 41N
2. ΣFy = may,
Ry2 − Ry1 − mg = may
Ry2 = 3 69 + 3 72 × 9 8 + 3 72 − 4 21 = 24 48N
3. About the COM of the leg, ΣM = Iα,

M 2 − M 1 − 0 169Rx1 + 0 185 Ry1 − 0 129 Rx2 + 0 142 Ry2 = Iα
M 2 = 1 34 + 0 169 × 10 52 − 0 185 × 3 69 + 0 129 × 10 41
− 0 142 × 24 48 + 0 0638 × 36 9
= 1 34 + 1 78 − 0 68 + 1 34 − 3 48 + 2 35 = 2 65 N m
Discussion
1. M2 is positive. This represents a counterclockwise (extensor) moment acting at the knee.
The quadricep muscles at this time are rapidly extending the swinging leg.
2. The angular acceleration of the leg is the net result of two reaction forces and one muscle
moment acting at each end of the segment. Thus, there may not be a single primary force
causing the movement we observe. In this case, each force and moment had a significant
influence on the final acceleration.
ISTUDY
5.2 FORCE TRANSDUCERS AND FORCE PLATES
In order to measure the force exerted by the body on an external body or load, we need a suitable
force-measuring device. Such a device, called a force transducer, gives an electrical signal pro-
portional to the applied force. There are many kinds available: strain gauge, piezoelectric, piezo-
resistive, capacitive, and others. All these work on the principle that the applied force causes a
certain amount of strain within the transducer. For the strain gauge type, a calibrated metal plate
or beam within the transducer undergoes a very small change (strain) in one of its dimensions.
This mechanical deflection, usually a fraction of 1%, causes a change in resistances connected as
a bridge circuit (see Chapter 3), resulting in an unbalance of voltages proportional to the force.
Piezoelectric and piezoresistive types require minute deformations of the atomic structure within
a block of special crystalline material. Quartz, for example, is a naturally found piezoelectrical
material, and deformation of its crystalline structure changes the electrical characteristics such
that the electrical charge across appropriate surfaces of the block is altered and can be translated
via suitable electronics to a signal proportional to the applied force. Piezoresistive types exhibit a
change in resistance which, like the strain gauge, upset the balance of a bridge circuit.
5.2.1 Multidirectional Force Transducers

In order to measure forces in two or more directions, it is necessary to use a bi- or tri-directional
force transducer. Such a device is nothing more than two or more force transducers mounted at
right angles to each other. The major problem is to ensure that the applied force acts through the
central axis of each of the individual transducers. As discussed previously there are several tech-
nologies used to measure force. The two most commonly used for commercial force plates are
strain gauges and piezoelectric crystals. Both technologies provide very accurate force measures
important for inverse solutions in biomechanics; however, it is important to note the advantages
and disadvantages of both technologies. This information will allow the user to identify the force
measuring technology most appropriate for the activities being studied.
Strain gauge force transducers

Advantages
• No drift over long data collections (i.e. postural stability)
• High sampling rate
Disadvantages
• Mid amplitude range
• Ballistic activities
Piezoelectric crystals force transducers

Advantages
• High amplitude range
• High sampling rate
• Ballistic activities
Disadvantages
• Drift over long data collections (i.e. postural stability)
ISTUDY
5.2 FORCE TRANSDUCERS AND FORCE PLATES 99
5.2.2 Force Plates

The most common force acting on the body is the ground reaction force, which acts on the foot
during standing, walking, or running. This force vector is three-dimensional (3D) and consists of
a vertical component plus two shear components acting along the force plate surface. These shear
forces are usually resolved into anterior–posterior and medial–lateral directions (Elftman, 1939).
The fourth variable needed is the location of the COP of this ground reaction vector. The foot
is supported over a varying surface area with different pressures at each part. Even if we knew
the individual pressures under every part of the foot, we would be faced with the expensive
problem of calculating the net effect of all these pressures as they change with time. It is, there-
fore, necessary to use a force plate to give us all the forces necessary for a complete inverse
solution.
Two common types of force plates are now explained. The first and most commonly used is a
flat plate supported by four triaxial transducers, depicted in Figure 5.9. Consider the coordinates
of each transducer to be (0, 0), (0, Z), (X, 0), and (X, Z). The location of the COP is determined by
the relative vertical forces seen at each of these corner transducers. If we designate the vertical
forces as F00, FX0, F0Z, and FXZ, the total vertical force is FY = F00 + FX0 + F0Z + FXZ. If all four
forces are equal, the COP is at the exact center of the force plate, at (X/2, Z/2). In general,
X F X0 + F XZ − F 00 + F 0Z
x= 1+ (5.4)
2 FY
Z F 0Z + F XZ − F 00 + F X0
z= 1+ (5.5)
2 FY
A second type of force plate and currently very rare has one centrally instrumented pillar that
supports an upper flat plate. Figure 5.10 shows the forces that act on this instrumented support.
The action force of the foot Fy acts downward, and the anterior-posterior shear force can act either
forward or backward. Consider a reverse shear force Fx, as shown. If we sum the moments acting
about the central axis of the support, we get:
M z − F y x + F x y0 = 0
F x y0 + M z (5.6)
x=
Fy
Foz F Fxz
(x1, z1)
Foo Fxo
Figure 5.9 Force plate with force transducers in the four corners. Magnitude and location of ground reaction force
F can be determined from the signals from the load cells in each of the support bases.
ISTUDY
Fy
Fx x
x
y0
Mz
Figure 5.10 Central support type force plate, showing the location of the center of pressure of the foot and the
forces and moments involved.
where
Mz = bending moment about axis of rotation of support
y0 = distance from support axis to force plate surface.
Since Fx, Fy, and Mz continuously change with time, x can be calculated to show how the COP
moves across the force plate. Caution is required for both types of force platforms when Fy is very
low (<2% of body weight). During this time, a small error in Fy represents a large percentage error
in x and z, as calculated by Equations (5.4), (5.5), and (5.6).
Typical force plate data are shown in Figure 5.11 plotted against time for a subject walking at a
normal speed. The vertical reaction force Fy is very characteristic in that it shows a rapid rise at
heel contact to a value in excess of body weight as full weight bearing takes place (i.e. the body
mass is being accelerated upward). Then as the knee flexes during midstance, the plate is partially
unloaded and Fy drops below body weight. At push off, the plantarflexors are active in causing
ankle plantarflexion. This causes a second peak greater than body weight. Finally, the weight
drops to zero as the opposite limb takes up the body weight. The anterior-posterior reaction force
Fx is the horizontal force exerted by the force plate on the foot. Immediately after heel contact, it is
negative, indicating a backward horizontal friction force between the floor and the shoe. If this
force were not present, the foot would slide forward, as sometimes happens when walking on icy
or slippery surfaces. Near midstance, Fx goes positive, indicating that the force plate reaction is
acting forward as the muscle forces (mainly plantarflexor) cause the foot to push back against
the plate.
The COP typically starts at the lateral heel, assuming that initial contact is made by the heel,
and then progresses forward toward the metatarsal heads and the first toe. The position of the
COP relative to the foot cannot be obtained from the force plate data themselves. We must first
know where the foot is, relative to the midline of the plate. The type of force plate used to collect
these data is the centrally instrumented pillar type depicted in Figure 5.10. It should be noted that
Mz is positive at heel contact and then becomes negative as the bodyweight moves forward. The
centers of pressure Xcp and Ycp are calculated in absolute coordinates to match those given in the
kinematics listing. Ycp was set at 0 to indicate ground level.
ISTUDY
100% B.W.
Fy
Mz
1s
Fx 200 N
Figure 5.11 Force plate record obtained during gait, using a central support type as shown in Figure 5.10.
It is very important to realize that these reaction forces are merely the algebraic summation of
all mass-acceleration products of all body segments. In other words, for an N-segment system the
reaction force in the x-direction is:
N
Fx = i=1
mi axi (5.7)
where
mi = mass of ith segment
axi = acceleration of ith-segment COM in the x-direction.
Similarly, in the vertical direction,
N
Fy = i=1
mi ayi + g (5.8)
where
ayi = acceleration of ith-segment COM in the y-direction
g = acceleration due to gravity.
With a person standing perfectly still on the platform, Fy will be nothing more than the
sum of all segment masses times g, which is the body weight. The interpretation of the
ISTUDY
ground reaction forces as far as what individual segments are doing is virtually impossible. In
the algebraic summation of mass-acceleration products, there can be many cancellations. For
example, FY could remain constant as one arm is accelerating upward, while the other has an
equal and opposite downward acceleration. Thus, to interpret the complete meaning of any
reaction force, we would be forced to determine the accelerations of the COMs of all segments
and carry out the summations seen in Equations (5.7) and (5.8).
5.2.2.1 Alternative Force Collection Devices

5.2.2.1.1 Instrumented Treadmills The implementation of instrumented treadmills has helped
to alleviate many of the limitations present when collecting walking and running biomechanical
data. First, many biomechanical labs are often small, which will not allow for an adequate amount
of strides to produce a natural gait pattern especially with running. Next, even when enough lab
space is available for walking or running unnatural strides often occur when attempting to target
the force plate. Lastly, treadmills allow researchers to control walking and running speed, which
can influence kinematics and kinetics.
There are two main styles of instrumented treadmills: (1) side-by-side treadmills (Figure 5.12)
and (2) tandem treadmills. Side-by-side treadmills are designed to provide consistent sagittal
plane kinematics but slight alterations in frontal plane kinematics may occur attempting to posi-
tion each foot on the belt. Side-by-side treadmills also provide researchers with the ability to inde-
pendently change speeds to create random perturbations or when testing pathological gait as seen
in stroke patients. Tandem treadmills are designed to maintain consistent kinematics and prevent
the frontal plane alteration seen in side-by-side treadmills. However, having the belts split front to
back can slightly alter sagittal plane kinematics due to maintaining the body’s COM at the tread-
mill split. This design also avoids force recording problems that occur during double limb support
during walking. Tandem treadmills are able to alter belt speed but are not typically done
independently.
The technical challenge of instrumented treadmills is the numerous external loads placed on
the treadmill during walking and running that have the potential to cause inaccurate ground reac-
tion force data. To avoid these errors instrumented treadmills have been designed to have the
Figure 5.12 Example of a Bertec instrumented side-by-side treadmill. (Photo provided with permission from
Fernando Santos of Bertec LLC.)
ISTUDY
entire treadmill mounted on the force transducers so that these external loads essentially cancel
each other out. The external load that is of most concern is the frictional force between the belt
and the surface of the force plate caused by the motors pulling the belt. Willems and Gosseye
(2013) used a model to demonstrate all forces that are applied to the treadmill during motion
and how these external forces are nil or cancel each other out including the friction between
the moving belt and the treadmill surface. As they reported, mechanical vibration occurs when
a participant is walking or running on the treadmill as the connections between components and
the material itself is not completely ridged. In addition, the deflections of the material can also
store and release energy altering force readings and accelerations of the participant. The vibra-
tions of the treadmills COM will introduce noise into the force readings. In addition, it will also
alter the signal-to-noise ratio, which ultimately can decrease accuracy in the COP calculation.
One solution for overcoming the noise is to average or integrate ground reaction forces over sev-
eral strides; however, this removes the ability to examine peak forces. Another solution is to use a
low pass filter to remove the vibration noise. Commercially available instrumented treadmills are
designed to have high natural frequencies so that the frequencies of the participant walking or
running don’t converge. This design feature minimizes the vibration noise and also allows the
use of a low pass filter to eliminate the high frequency vibration noise. In light of some of
the technical challenges, several studies have demonstrated accurate and reliable data collected
on instrumented treadmills as compared to overground walking and running.
5.2.2.1.2 Instrumented Handrails Instrumented handrails have been integrated into instru-
mented treadmills as a way to calculate both lower and upper extremity inverse dynamics in
populations that require gait assistance either unilaterally or bilaterally. Instrumented handrails
have also been used to examine inverse dynamics during stair ascent and descent. The handrails
utilize standard three-component force transducers. Therefore, the mathematical approach is the
same with the exception of correcting for the angle of the handrail when mounted on instrumen-
ted stairs or when treadmills are in inclined positions. However, most commercially available
treadmills that incline will account for this within their collection software.
5.2.2.1.3 Insole Pressure Sensors The COP measured by the force plate is a weighted average
of the distributed COPs under the two feet if both are in contact with the ground or under the one
foot that is in contact. However, the COP signal tells us nothing about the pressure at any of the
contact points under the foot. For example, during midstance in walking or running, there are two
main pressure areas: the ball of the foot and the heel, but the force plate records the COP as being
under the arch of the foot, where there may in fact be negligible pressure. To get some insight into
the distributed pressures around all contact points, a number of special pressure measurement
systems have been developed. A typical pressure measurement system was introduced a number
of years ago by Tekscan and is schematically shown in Figure 5.13. It is a flexible sheet of tactile
force sensors that is cut in the shape of a shoe insole and that can be trimmed down to any shoe
size. Each sensel shown in the exploded view is about 0.2 in. × 0.2 in. (5 mm × 5 mm), giving
25 sensels/sq. in. Two thin flexible polyester sheets have electrically conductive electrodes
deposited in rows and columns. A thin semiconductive coating is applied between the conductive
rows and columns, and its electrical resistance changes with the pressure applied. The matrix of
sensels results in a matrix of voltage changes that are computer displayed as different colors for
different pressure levels. Colors range from blue through green, yellow, orange, and red (1–125
psi). Thus, the high- and low-pressure areas under the foot are visually evident as the pressure
moves from the heel to the toes over the stance period (Hsiao et al., 2002). Such devices have
been valuable in identifying high-pressure points in various foot deformities and in diabetic feet
(Pitie et al., 1999). The results of surgery and shoe orthotics to relieve the pressure points are also
immediately available. Lastly, these sensors have greatly enhanced the development of footwear
(including various forms of athletic footwear) and
ISTUDY
Sensor model number: 3000/3001

Sensor name: F-SCAN
Column width (CW)
Row spacing (RS)
Row width (RW) Sensel
Column spacing (CS)

Tab length (A)
Exploded view
Overall length (L)
Matrix height (MH)
Matrix width (MW)

Overall width (W)
Figure 5.13 Foot pressure measurement system introduced by Tekscan, Inc. It is a shoe insole insert with an array
of pressure sensors that vary their resistance with applied pressure. See the text for detailed operation of sensor and
computer display system. (Reproduced with permission of Tekscan, Inc.)
5.2.2.1.4 Instrumented Walking Boots An instrumented walking boot was designed (Hullfish
et al., 2020) and used to estimate Achilles tendon load. The boot was outfitted with an insole
pressure sensor and a load cell mounted to the posterior strut of the walking boot to calculate
the overall resistance moment created from the boot. In this study, the instrumented walking boot
was used in conjunction with a motion capture system including a force plate. The motion capture
system was used to both validate the walking boot and was also required to calibrate the insole
sensor. Therefore, in its current form, the walking boot is not able to be used outside of a motion
capture lab. Nonetheless, this walking boot can provide useful information specifically to Achil-
les tendon loading. It is important to note that this experimental approach has several assump-
tions: (1) the plantar loads measured with the insole sensor are all created from the plantar
flexor muscles of the ankle and passed directly through the Achilles tendon and (2) the posterior
strut of the boot resists all of the reaction loads during walking. It is currently unknown if these
assumptions are true as the toe flexor muscles can create plantar forces and there are also altered
neuromuscular strategies in patients with pain or injury. However, the current approach was
found to detect decreases in Achilles tendon loads that occur when the ankle is positioned in
greater amounts of plantarflexion within the boot.
5.2.3 Combined Force Plate and Kinematic Data

It is valuable to see how the reaction force data from the force plate are combined with the
segment kinematics to calculate the muscle moments and reaction forces at the ankle joint during
dynamic stance. This is best illustrated in a sample calculation. For a subject in late stance (see
Figure 5.14), during toe-off the following foot accelerations were recorded: ax = 3.25 m/s2,
ISTUDY
Fay Frame 35
Ma
Fy = 765.96 N
(89.7, 16.0)
a Fax
(95.3, 8.4)
1.12 g (100.9, 0.9)

h Fx = 160.25 N
m–p
Ground (103.2, 0)
Figure 5.14 Free-body diagram of foot during weight bearing with the ground reaction forces, Fx and Fy, shown as
being located at the COP.
ay = 1.78 m/s2, and α = − 45.35 rad/s2. The mass of the foot is 1.12 kg, and the moment of inertia
is 0.01 kg m2.
Example 5.4 (see Figure 5.14). From Equation (5.1),
F ax + F x = max
F ax = 1 12 × 3 25 − 160 25 = − 156 6 N
From Equation (5.2),
F ay + F y − mg = may
F ay = 1 12 × 1 78 − 765 96 + 1 12 × 9 81 = − 753 0 N
From Equation (5.3), about the COM of the foot, ΣM = Iα,
M a + F x × 0 084 + F y × 0 079 − F ay × 0 056 − F ax × 0 076

= 0 01 − 45 35
M a = − 0 01 × 45 35 − 0 084 × 160 25 − 0 079 × 765 96 − 0 056
× 753 0 − 0 076 × 156 6 = − 128 5 N m
The polarity and the magnitude of this ankle moment indicate strong plantarflexor activity
acting to push off the foot and cause it to rotate clockwise about the metatarsophalangeal joint.
5.2.4 Interpretation of Moment-of-Force Curves

A complete link-segment analysis yields the net muscle moment at every joint during the time
course of movement (Pedotti, 1977). As an example, we discuss in detail the ankle moments-of-
force of a hip joint replacement patient, as shown in Figure 5.15. Three repeat trials are plotted.
ISTUDY
80
60
40 Ms Ms = Mk – Ma –Mh
20
–20
Flex.
–
20
Mh
Joint moments of force (N. m)
0 Mh
Ext.
–20
+
–40
Mk Mk
20
Ext.
0
–
–20
Flex.
Ma
–40
20
0 WP45D
WP45F
Plantor.
–20
Ma WP45G
–40
TO
HC
–60
0 200 400 600 800
Time (ms)
Figure 5.15 Joint moment-of-force profiles from three repeat trials of a patient fitted with a total hip replacement.
See the text for a detailed discussion.
The convention for the moments is shown to the right of the plot. Counterclockwise moments
acting on a segment distal to the joint are positive; clockwise moments are negative. Thus, a plan-
tarflexor moment (acting on its distal segment) is negative, a knee extensor moment is shown to
be positive, and a hip extensor moment is negative.
The moments are plotted during stance, with heel contact at 0 and toe-off at 680 ms. The ankle
muscles generate a positive (dorsiflexor) moment for the first 80 ms of stance as the anterior tibial
muscles act eccentrically to lower the foot to the ground. Then the plantarflexors increase their
activity during mid- and late stance. During midstance, they act to control the amount of forward
rotation of the leg over the foot, which is flat on the ground. As the plantarflexors generate their
peak of about 60 N m, they cause the foot to plantarflex and create toe-off. This concentric action
results in a major generation of energy in healthy patients (Winter, 1983), but in this patient, it
was somewhat reduced because of the pathology related to the patient’s hip joint replacement.
Prior to toe-off, the plantarflexor moment drops to 0 because that limb is now unloaded due to the
fact that the patient is now weight bearing on their good limb and, for the last 90 ms prior to toe-
off, the toe is just touching the ground with a light force. At this same point in time, the hip flexors
are active to pull this limb upward and forward as a first phase of lower limb swing.
The knee muscles effectively show one pattern during all of stance. The quadriceps are active
to generate an extensor moment, which acts to control the
ISTUDY
stance and also extends the knee during midstance. Even during toe-off, when the knee starts
flexing in preparation for swing, the quadriceps act eccentrically to control the amount of knee
flexion. At heel contact, the hip moment is negative (extensor) and remains so until midstance.
Such activity has two functions. First, the hip muscles act on the thigh to assist the quadriceps in
controlling the amount of knee flexion. Second, the hip extensors act to control the forward rota-
tion of the upper body as it attempts to rotate forward over the hip joint (the reaction force at the
hip has a backward component during the first half of stance). Then, during the latter half of
stance, the hip moment becomes positive (flexor), initially to reverse the backward rotating thigh
and then, as described earlier, to pull the thigh forward and upward.
The fourth curve, Ms, bears some explanation. It is the net summation of the moments at all
three joints, such that extensor moments are positive. It is called the support moment (Winter,
1980) because it represents a total limb pattern to push away from the ground. In scores of walk-
ing and running trials on healthy participants and patients, this synergy has been shown to be
consistently positive during single support, in spite of considerable variability at individual joints
(Winter, 1984). This latter curve is presented as an example to demonstrate that moment-of-force
curves should not be looked at in isolation but rather as part of a total integrated synergy in a given
movement task. A complete discussion and analysis of this synergy are presented in Chapter 11.
5.2.5 Differences Between Center of Mass and Center of Pressure

For many in the applied and clinical areas, the terms COM and COP are often misinterpreted or
even interchanged. The COM of the body is the net location of the COM in 3D space and is the
weighted average of the COM of each segment as calculated in Chapter 4. The location of the
COM in the vertical direction is sometimes called the center of gravity (COG). The trajectory of
this vertical line from the COM to the ground allows us to compare the trajectories of the COM
and COP. The trajectory of the COP is totally independent of the COM, and it is the location of
the vertical ground reaction force vector from a single force platform, assuming that all body
contact points are on that platform. The vertical ground reaction force is a weighted average
of the location of all downward (action) forces acting on the force plate. These forces depend
on the foot placement and the motor control of the ankle muscles. Thus, the COP is the neuro-
muscular response to the imbalances of the body’s COM. The major misuse of the COP comes
from researchers who refer to COP as “sway,” thereby inferring it to be the kinematic meas-
ure COM.
The difference between COM and COP is demonstrated in Figure 5.16. Here, we see a subject
swaying back and forth while standing erect on a force plate. Each figure shows the changing
situation at one of five different points in time. Time 1 has the body’s COM (shown by the vertical
bodyweight vector, W) to be ahead of the COP (shown by the vertical ground reaction vector, R).
This “parallelogram of forces” acts at distances g and p, respectively, from the ankle joint. The
magnitudes of W and R are equal and constant during quiet standing. Assuming that the body to
be pivoting about the ankles and neglecting the small mass of the feet, a counterclockwise
moment equal to Rp and a clockwise moment equal to Wg will be acting. At Time 1, Wg >
Rp and the body will experience a clockwise angular acceleration, α. It will also have a clockwise
angular velocity, ω. In order to correct this forward imbalance, the subject will increase their plan-
tarflexor activity, which will increase the COP such that at Time 2 the COP will be anterior of the
COM. Now Rp > Wg. Thus, α will reverse and will start to decrease ω until, at Time 3, the time
integral of α will result in a reversal of ω. Now both ω and α are counterclockwise, and the body
will be experiencing a backward sway. The subject’s response to the backward sway at Time 4 is
to decrease their COP by reduced plantarflexor activation. Now Wg > Rp and α will reverse, and,
after a period of time, ω will decrease and reverse, and the body will return to the original con-
ditions, as seen at Time 5. From this sequence of events relating COP to COM, it is evident that
the plantarflexors/dorsiflexors vary the net ankle moment to control the COP and thereby regulate
the body’s COM. However, it is apparent that the COP must
ISTUDY
ω ω
ω ω ω
α α α α α
R R R R R
g g g g g
W W
W W W
p p p p p
1 2 3 4 5
Figure 5.16 A subject swaying back and forth while standing on a force platform. Five different points in time are
described, showing the COM and the COP locations along with the associated angular accelerations and velocities of
the body. See the text for a detailed description.
posterior of the COM; thus, the dynamic range of the COP must be somewhat greater than that of
the COM. If the COM were allowed to move within a few centimeters of the toes, it is possible
that a corrective movement of the COP to the extremes of the toes might not be adequate to
reverse ω. Here, the subject would be forced to take a step forward to arrest a forward fall.
Figure 5.17 is a typical 40-second record of the center of pressure (COPx) and center of mass
(COMx) in the anterior/posterior direction of an adult subject standing quietly. Note that both
signals are virtually in phase and that COP is slightly greater than COM. As was seen in the dis-
cussion regarding Figure 5.16, the COP must move ahead of and behind the COM in order to
decelerate it and reverse its direction. Note that all reversals of direction of COM coincide with
an overshoot of the COP signal.
5.2.6 Kinematics and Kinetics of the Inverted Pendulum Model

All human movement (except in space) is done in a gravitational environment, and, therefore,
posture and balance are continuous tasks that must be accomplished. In normal daily activity
at home, at work, and in our sports and recreation, we must maintain a safe posture and balance.
The base of support can vary from one foot (running) to a four-point support (football), and it is
essential that the COM remain within that base of support or move safely between the two feet if it
lies temporarily outside the base of support (as it does in running and during the single-support
phase of walking). There is a common model that allows us to analyze the dynamics of balance:
the inverted pendulum model, which relates the trajectories of the COP and COM. As was seen in
ISTUDY
COP, COM in anterior/posterior direction (50% stance width)
0.8 COPx
COMx
0.4
Displacement (cm)
–0.4
–0.8
–1.2
0 5 10 15 20 25 30 35 40
Time (s)
Figure 5.17 Typical 40-second record of the total body center of mass (COMx) and center of pressure (COPx) in
the anterior/posterior direction during quiet standing. The COP amplitude exceeds that of the COM, and the
reversals of direction of the COM (i.e. high + ve or − ve accelerations) are caused by overshoots of COP, as
predicted by the discussion related to Figure 5.16.
Section 5.2.5, the position of the COP relative to the COM decides the direction of the angular
acceleration of the inverted pendulum. A full biomechanical analysis of the inverted pendulum
model in both sagittal and frontal planes has been presented by Winter et al. (1998). In the sagittal
plane, assuming that the body swayed about the ankles, it was shown that:
I S COMx
COPx − COMx = − (5.9)
Wh
where
Is = moment of inertia of body about ankles in sagittal plane
CÖMx = forward acceleration of COM
W = body weight above ankles
h = height of COM above ankles.
In the frontal plane, the balance equation is virtually the same:
I f COMz
COPz − COMz = − (5.10)
Wh
where
If = moment of inertia of body about ankles in frontal plane
CÖMz = medial/lateral acceleration of COM.
The major difference between Equations (5.9) and (5.10) is in the muscle groups that
control the COP. In Equation (5.9), COPx = Ma/R, where Ma is the sum of the right and
left plantarflexor moments and R is the total vertical reaction force at the ankles. In Equa-
tion (5.10), COPz = Mt/R, where Mt = Mal + Mar + Mhl + Mhr. Mal with Mar are the left and
right frontal ankle moments, while Mhl and Mhr are the left and right frontal plane hip
ISTUDY
moments. Thus, frontal plane balance is controlled by four torque motors, one at each corner
of the closed-link parallelogram consisting of the two lower limbs and the pelvis. The hip
abductor/adductor moments have been shown to be totally dominant in side-by-side stand-
ing (Winter et al., 1996), while the ankle invertor/evertor moments play a negligible role in
balance control.
The validity of the inverted pendulum model is evident in the validity of Equations (5.9) and
(5.10). As COP and COM are totally independent measures, then the correlation of (COP–COM)
with CÖM will be a measure of the validity of this simplified model. Validations of the model
during quiet standing (Gage et al., 2004) showed a correlation of r = −0.954 in the anterior/pos-
terior (A/P) direction and r = −0.85 in the medial/lateral (M/L) direction. The lower correlation in
the M/L direction was due to the fact that the M/L COM displacement was about 45% of that in
the A/P direction. Similar validations have been made to justify the inverted pendulum model
during initiation and termination of gait (Jian et al., 1993); correlations averaged −0.94 in both
A/P and M/L directions.
5.3 BONE-ON-BONE FORCES DURING DYNAMIC CONDITIONS
Link-segment models assume that each joint is a hinge joint and that the moment of force is gen-
erated by a torque motor. In such a model, the reaction force calculated at each joint would be the
same as the force across the surface of the hinge joint (i.e. the bone-on-bone forces). However, our
muscles are not torque motors; rather, they are linear motors that produce additional compressive
and shear forces across the joint surfaces. Thus, we must overlay on the free-body diagram these
additional muscle-induced forces. In the extreme range of joint movement, we would also have to
consider the forces from the ligaments and anatomical constraints. However, for the purposes of
this text, we will limit the analyses to estimated muscle forces.
5.3.1 Indeterminacy in Muscle Force Estimates

Estimating muscle force is a major problem, even if we have good estimates of the moment of
force at each joint. The solution is indeterminate. Figure 5.18 demonstrates the number of major
muscles responsible for the sagittal plane joint moments of force in the lower limb. At the knee,
for example, there are nine muscles whose forces create the net moment from our inverse solu-
tion. The line of action of each of these muscles is different and continuously changes with time.
Thus, the moment arms are also dynamic variables. Therefore, the extensor moment is a
net algebraic sum of the cross product of all force vectors and moment arm vectors,
Ne Nf
Mj t = F t × d ei t −
i = 1 ei
F
i = 1 fi
t × d fi t (5.11)
where
Ne = number of extensor muscles
Nf = number of flexor muscles
Fei(t) = force in ith extensor muscle at time t
dei(t) = moment arm of ith extensor muscle at time t.
Thus, a first major step is to make valid estimates of individual muscle forces and to combine
them with a detailed kinematic/anatomical model of lines of pull of each muscle relative to the
joint’s center (or our best estimate of it). Thus, a separate model must be developed for each joint,
and a number of simplifying assumptions are necessary in order to resolve the indeterminacy
problem. An example is now presented of a runner during the rapid toe-off phase when the plan-
tarflexors are dominant.
ISTUDY
5.3 BONE-ON-BONE FORCES DURING DYNAMIC CONDITIONS 111
Major muscles Moments of force
Iliopsoas
Gluteus maximus
Semitendinosus
Semimembranosus Mh
Biceps femoris
–
Sartorius
Mh
Rectus femoris
Vastus medialis Mk Knee angle
Vastus lateralis +
Vastus intermedius Mk
Gastrocnemius
Soleus
–
Tibialis posterior Ma
Ma
Peronei
Tibialis anterior
Figure 5.18 Fifteen major muscles are responsible for the sagittal plane moments of force at the ankle, knee, and
hip joints. During weight bearing, all three moments control the knee angle. Thus, there is considerable
indeterminacy when relating knee angle changes to any single moment pattern or to any unique combination of
muscle activity.
5.3.2 Example Problem

During late stance, a runner’s foot and ankle are shown (see Figure 5.19) along with the direction of
pull of each of the plantarflexors. The indeterminacy problem is solved by assuming that there is no
co-contraction and that each active muscle’s stress is equal [i.e. its force is proportional to its phys-
iological cross-sectional area (PCA)]. Thus, Equation (5.11) can be modified as follows for the five
major muscles of the extensor group acting at the ankle (Scott and Winter, 1990):
5
Ma t = i=1
PCAi × Sei t × dei t (5.12)
Since the PCA for each muscle is known and dei can be calculated for each point in time, the
stress, Sei(t) can be estimated.
For this model, the ankle joint center is assumed to remain fixed and is located as shown in
Figure 5.20. The location of each muscle origin and insertion is defined from the anatomical mar-
kers on each segment, using polar coordinates. The attachment point of the ith muscle is defined
by a distance Ri from the joint center and an angle θmi between the segment’s neutral axis and the
line joining the joint center to the attachment point. Thus, the coordinates of any origin or inser-
tion at any instant in time are given by:
X mi t = X i t + Ri cos θmi t + θs t (5.13a)

Y mi t = Y i t + Ri sin θmi t + θs t (5.13b)
where θs(t) is the angle of the foot segment in the spatial coordinate system. In Figure 5.20, the
free-body diagram of the foot is presented, showing the internal anatomy to demonstrate the pro-
blems of defining the effective insertion point of the muscles and the effective line of pull of each
muscle. Four muscle forces are shown here: soleus Fs, gastrocnemius Fg, flexor hallucis longus
Fh, and peronei Fp. The ankle joint’s center is defined by
ISTUDY
Fa
Fp Fd + Ft
Ra
Fh
Fa - Soleus + gastrocnemius
Fh - Hallucis longus Rg
Fp - Peronei (brevis + longus)
Fd - Digitorum longus
Ft - Tibialis posterior
Ra - Tibial reaction force
Rg - Ground reaction force
Figure 5.19 Anatomical drawing of foot and ankle during the toe-off phase of a runner. The tendons for the major
plantarflexors as they cross the ankle joint are shown, along with ankle and ground reaction forces.
Fg
F′p F′h
Fh
Fs
Fp
Ra
β
R
θs
θm
Rg
Figure 5.20 Free-body diagram of foot segment showing the actual and effective lines of pull of four of the
plantarflexor muscles. Also shown are the reaction forces at the ankle
ISTUDY
5.3 BONE-ON-BONE FORCES DURING DYNAMIC CONDITIONS 113
The insertion of the Achilles tendon is at a distance R with an angle θm from the foot segment
(defined by a line joining the ankle to the fifth-metatarsal–phalangeal joint). The foot angular
position is θs in the plane of movement. The angle of pull for the soleus and gastrocnemius mus-
cles from this insertion is rather straightforward. However, for Fh and Fp the situation is quite
different. The effective angle of pull is the direction of the muscle force as it leaves the foot seg-
ment. The flexor hallucis longus tendon curves under the talus and inserts on the distal phalanx of
the big toe. As this tendon leaves the foot, it is rounding the pulleylike groove in the talus. Thus,
its effective direction of pull is F h. Similarly, the peronei tendon curves around the distal end of
the lateral malleolus. However, its effective direction of pull is F p .
The moment arm length dei for any muscle required by Equation (5.12) can now be calculated:
dei = Ri sin βi (5.14)
where βi is the angle between the effective direction of pull and the line joining the insertion point
to the joint center.
In Figure 5.20, β for the soleus only is shown. Thus, it is now possible to calculate Sei(t) over
the time that the plantarflexors act during the stance phase of a running cycle. Thus, each muscle
force Fei(t) can be estimated by multiplying Sei(t) by each PCAi. With all five muscle forces
known, along with Rg and Ra, it is possible to estimate the total compressive and shear forces
acting at the ankle joint. Figure 5.21 plots these forces for a middle-distance runner over the
stance period of 0.22 s. The compressive forces reach a peak of more than 5500 N, which is
6000 Compressive
5000 Total
4000
3000 Muscle
2000
Reaction
1000 B.W.
Force (N)
0
600
400 Muscle Shear
Ant
200
0
–200 Total
–400
Post
–600 Reaction
–800
0.00
0.02
0.04
0.06
0.08
0.10
0.12
0.14
0.16
0.18
0.20
Stance time (s)
Figure 5.21 Compressive and shear forces at the ankle calculated during the stance phase of a middle-distance
runner. The reaction force accounts for less than 20% of the total compressive force. The direction of pull of the
major plantarflexors is such as to cause an anterior shear force (the talus is shearing anteriorly with respect to
the tibia), which is opposite to that caused by the reaction forces.
an anti-shear mechanism.
ISTUDY
in excess of 11 times this runner’s body weight. It is interesting to note that the ground reaction
force accounts for only 1000 N of the total force, but the muscle forces themselves account for
over 4500 N. The shear forces (at right angles to the long axis of the tibia) are actually reduced by
the direction of the muscle forces. The reaction shear force is such as to cause the talus to move
posteriorly under the tibia. However, the line of pull of the major plantarflexors (soleus and
gastrocnemius) is such as to pull the foot in an anterior direction. Thus, the total shear force
is drastically reduced from 800 N to about 300 N, and the muscle action has been classified
as an anti-shear mechanism.
5.4 REFERENCES
Bresler, B. and J. P. Frankel. “The Forces and Moments in the Leg during Level Walking,” Trans. ASME 72:
27–36, 1950.
Elftman, H. “Forces and Energy Changes in the Leg During Walking,” Am. J. Physiol. 125: 339–356, 1939.
Gage, W. G., D. A. Winter, J. S. Frank, and A. L. Adkin. “Kinematic and Kinetic Validation of Inverted
Pendulum Model in Quiet Standing,” Gait Posture 19: 124–132, 2004.
Hsiao, H., J. Guan, and M. Wetherly. “Accuracy and Precision of Two In-Shoe Pressure Measurement
Systems,” Ergonomics 45: 537–555, 2002.
Hullfish, T. J., K. M. O’Connor, and J. R. Baxter. “Instrumented Immobilizing Boot Paradigm Quantifies
Reduced Achilles Tendon Loading During Gait,” J. Biomech. 26: 109925, 2020.
Paul, J. P. “Forces Transmitted by Joints in the Human Body,” Proc. Inst. Mech. Eng. 18(3): 8–15, 1966.
Pedotti, A. “A Study of Motor Coordination and Neuromuscular Activities in Human Locomotion,” Biol.
Cybern. 26: 53–62, 1977.
Pitie, D. L., M. Lord, A. Foster, S. Wilson, P. Watkins, and M. E. Edmonds. “Plantar Pressures are Elevated in
the Neuroischemic and Neuropathic Diabetic Foot,” Diabetic Care 23: 1966–1970, 1999.
Scott, S. H. and D. A. Winter. “Internal Forces at Chronic Running Injury Sites,” Med. Sci. Sports Exercise 22:
357–369, 1990.
Willems, P. A. and T. P. Gosseye. “Does an Instrumented Treadmill Correctly Measure the Ground Reaction
Forces?” Biol. Open 2 (12): 1421–1424, 2013.
Winter, D. A. “Overall Principle of Lower Limb Support during Stance Phase of Gait,” J. Biomech. 13:
923–927, 1980.
Winter, D. A. “Energy Generation and Absorption at the Ankle and Knee during Fast, Natural and Slow
Cadences,” Clin. Orthop. Rel. Res. 197: 147–154, 1983.
Winter, D. A. “Kinematic and Kinetic Patterns in Human Gait: Variability and Compensating Effects,” Hum.
Mov. Sci. 3: 51–76, 1984.
Winter, D. A., F. Prince, J. S. Frank, C. Powell, and K. F. Zabjek. “A Unified Theory Regarding A/P and M/L
Balance During Quiet Standing,” J. Neurophysiol. 75: 2334–2343, 1996.
Winter, D. A., A. E. Patla, F. Prince, M. G. Ishac, and K. Gielo-Perczak. “Stiffness Control of Balance in Quiet
ISTUDY
6
MECHANICAL WORK, ENERGY,
AND POWER
1
2
3
6.0 INTRODUCTION
If we were to choose which biomechanical variable contains the most information, we would be
forced to look at a variable that relates to the energetics. Without that knowledge, we would know
nothing about the energy flows that cause the movement we are observing; and no movement
would take place without those flows. Diagnostically, we have found joint mechanical powers
to be the most discriminating in all our assessment of pathological gait. Without them, we could
have made erroneous or incomplete assessments that would not have been detected by electro-
myographic (EMG) or moment-of-force analyzes alone. Also, valid mechanical work calcula-
tions are essential to any efficiency assessments that are made in sports and work-related tasks.
Before proceeding, the student should have in mind certain terms and laws relating to mechan-
ical energy, work, and power, and these will now be reviewed.
6.0.1 Mechanical Energy and Work

Mechanical energy and work have the same units (joules) but have different meanings. Mechan-
ical energy is a measure of the state of a body at an instant in time as to its ability to do work. For
example, a body which has 200 J of kinetic energy (KE) and 150 J of potential energy (PE) is
capable of doing 350 J of work (on another body). Work, on the other hand, is the measure of
energy flow from one body to another, and time must elapse for that work to be done. If energy
flows from body A to body B, we say that body A does work on body B; or muscle A can do work
on segment B if energy flows from the muscle to the segment.
†
Deceased.

Companion website:
ISTUDY
116 MECHANICAL WORK, ENERGY, AND POWER
(a)
4J 2.4 J 3.8 J
5.3 J 1.7 J 0.2 J
ΔEs = 6.0 J
(b)
200 W 120 W 190 W

265 W 85 W 10 W
ΔEs
= 300 W
Δt
Figure 6.1 (a) Flow of energy into and out of a segment from the adjacent connective tissue and joint contacts over
a period of time Δt. (b) Rate of flow of energy (power) for the same segment and same point in time as (a). A “power
balance” can be calculated; see the text for discussion.
6.0.2 Law of Conservation of Energy

At all points in the body at all instants in time, the law of conservation of energy applies. For
example, any body segment will change its energy only if there is a flow of energy or out of
any adjacent structure (tendons, ligaments, or joint contact surfaces). Figure 6.1 depicts a segment
that is in contact at the proximal and distal ends and has four muscle attachments. In this situation,
there are six possible routes for energy flow. Figure 6.1a shows the work (in joules) done at each
of these points over a short period of time Δt. The law of conservation of energy states that the
algebraic sum of all the energy flows must equal the energy change of that segment. In the case
shown, ΔE s = 4 + 2 4 + 5 3 − 1 7 − 0 2 − 3 8 = 6 0 J. Thus, if we are able to calculate each of
the individual energy flows at the six attachment points, we should be able to confirm ΔEs
through an independent analysis of mechanical energy of that segment. The balance will not
be perfect because of measurement errors and because our link-segment model does not perfectly
satisfy the assumptions inherent in link-segment analyzes. A second way to look at the energy
balance is through a power balance, which is shown in Figure 6.1b and effectively looks at the-
rate of flow of energy into and out of the segment and equates that to the rate of change of energy
of the segment. Thus if, Δt is 20 ms, ΔE s Δt = 200 + 120 + 265 − 85 − 10 − 190 = 300 W.
Another aspect of energy conservation takes place within each segment. Energy storage within
each segment takes the form of potential and KE (translational and rotational). Thus, the segment
energy Es at any given point in time could be made up of any combination of potential and kinetic
energies, quite independent of the energy flows into or out of the segment. In Section 6.3.1, the
analysis of these components and the determination of the amount of conservation that occurs
within a segment over any given period of time are demonstrated.
6.0.3 Internal Versus External Work

The only source of mechanical energy generation in the human body is the muscles, and the major
site of energy absorption is also the muscles. A very small fraction of energy is dissipated into
heat as a result of joint friction and viscosity in the connective tissue. Thus, mechanical energy is
continuously flowing into and out of muscles and from segment to segment. To reach an external
load, there may be many energy changes in the intervening segments between the source and the
external load. In a lifting task (see Figure 6.2), the work rate
ISTUDY
6.0 INTRODUCTION 117
dEb
= 400 W
dt
200 W
100 W 50 W
300 W
50 W
100 W
Figure 6.2 Lifting task showing the power generation from a number of muscles and the combined rate of change
of energy of the body (internal work) and rate of energy flow to the load (external work).
but the work rate to increase the energy of the total body by the source muscles of the lower limb
might be 400 W. Thus, the sum of the internal and external work rates would be 600 W, and this
generation of energy might result from many source muscles, as shown. Or, during many move-
ment tasks such as walking and running, there is no external load, and all the energy generation
and absorption are required simply to move the body segments themselves. A distinction is made
between the work done on the body segments (called internal work) and the work done on the
load (called external work). Thus, lifting weights, pushing a car, or cycling an ergometer have
well-defined external loads. One exception to external work definition includes lifting one’s own
body weight to a new height. Thus, running up a hill involves both internal and external work.
External work can be negative if an external force is exerted on the body and the body gives way.
Thus, in contact sports, external work is regularly done on players being pushed or tackled.
A baseball does work on the catcher as his hand and arms give way.
In bicycle ergometry, the cyclist does internal work just to move his limbs through the cycle
(freewheeling). Figure 6.3 shows a situation where the cyclist did both internal and external work.
This complex experiment has one bicycle ergometer connected via the chain to a second bicycle.
Thus, one cyclist can bicycle in the forward direction (positive work), while the other cycles
backward (negative work). The assumption made by the researchers who introduced this novel
idea was that each cyclist was doing equal amounts of work (Abbot et al., 1952). This is not true
because the positive-work cyclist must do his or her own internal work plus the internal work on
the negative-work cyclist plus any additional negative work of that cyclist. Thus, if the internal
work of each cyclist were 75 W, the positive-work cyclist would have to do mechanical work at
150 W rate just to “freewheel” both cyclists. Then as the negative-work cyclist contracted his
or her muscles, an additional load would be added. Thus, if the negative-work cyclist worked
at 150 W, the positive-work cyclists would be loaded to 300 W. It is no wonder that the nega-
tive-work cyclist can cycle with ease while the positive-work cyclist rapidly fatigues. They
are simply not working at the same mechanical work rate,
work far exceeds that of negative work.
ISTUDY
15
0 W
W
300
225 W
Figure 6.3 Bicycle ergometry situation in which one subject (left) cycles in a forward direction and does work on a
second subject who cycles in the reverse direction. The positive-work cyclist not only does external work on the
negative-work cyclist but also does the internal work to move the limbs of both cyclists. Contrary to common
interpretation, both cyclists are not performing equal magnitudes of mechanical work.
6.0.4 Positive Work of Muscles

Positive work is work done during a concentric contraction, when the muscle moment acts in the
same direction as the angular velocity of the joint. If a flexor muscle is causing a shortening, we
can consider the flexor moment to be positive and the angular velocity to be positive. The product
of muscle moment and angular velocity is positive; thus, power is positive, as depicted in
Figure 6.4a. Conversely, if an extensor muscle moment is negative and an extensor angular veloc-
ity is negative, the product is still positive, as shown in Figure 6.4b. The integral of the power over
the time of the contraction is the net work done by the muscle and represents generated energy
transferred from the muscles to the limbs.
6.0.5 Negative Work of Muscles

Negative work is work done during an eccentric contraction when the muscle moment acts in the
opposite direction to the movement of the joint. This usually happens when an external force,
Fext, acts on the segment and is such that it creates a joint moment greater than the muscle
moment. The external force could include gravitational or ground reaction forces. Using the
polarity convention as described, we can see in Figure 6.5a that we have a flexor moment
(a) (b)
Triceps EMG Biceps EMG

Biceps EMG Triceps EMG
Mf
ωf
ωe
Me
Figure 6.4 Positive power as defined by the net muscle moment and angular velocity. (a) A flexion moment acts
while the forearm is flexing. (b) An extension moment acts during an
ISTUDY
(a) (b)
Triceps EMG Biceps EMG

Biceps EMG Triceps EMG
Mf Fext
ωf
ωe
Me Fext
Figure 6.5 Negative power as defined by net muscle moment and angular velocity. (a) An external force causes
extension when the flexors are active. (b) An external force causes flexion in the presence of an extensor muscle
moment.
(positive) with an extensor angular velocity (negative). The product yields a negative power, so
that the work done during this angular change is negative. Similarly, when there is an extensor
moment (negative) during a flexor angular change (positive), the product is negative
(Figure 6.5b). Here, the net work is being done by the external force on the muscles and repre-
sents a flow of energy from the limbs into the muscles (absorption).
6.0.6 Muscle Mechanical Power

The rate of work done by most muscles is rarely constant with time. Because of rapid time-course
changes, it has been necessary to calculate muscle power as a function of time (Elftman, 1939;
Quanbury et al., 1975; Cappozzo et al., 1976; Winter and Robertson, 1978). At a given joint,
muscle power is the product of the net muscle moment and angular velocity,
Pm = M j ωj W (6.1)
where
Pm = muscle power, W
Mj = net muscle moment, N m
ωj = joint angular velocity, rad/s.
As has been described in the previous sections, Pm can be either positive or negative. During
even the simplest movements, the power will reverse sign several times. Figure 6.6 depicts the
muscle moment, angular velocities, and muscle power as a function of time during a simple
extension and flexion of the forearm. As can be seen, the time courses of Mj and ωj are roughly
out of phase by 90 . During the initial extension, there is an extensor moment and an extensor
angular velocity as the triceps do positive work on the forearm. During the latter extension phase,
the forearm is decelerated by the biceps (flexor moment). Here the biceps are doing negative work
(absorbing mechanical energy). Once the forearm is stopped, it starts accelerating in a flexor
direction still under the moment created by the biceps, which are now doing positive work.
Finally, at the end of the movement, the triceps decelerate the forearm as the extensor muscles
lengthen; here, Pm is negative.
6.0.7 Mechanical Work of Muscles

Until now, we have used the terms power and work almost interchangeably. Power is the rate of
doing work. Thus, to calculate work done, we must
ISTUDY
Ex.
Mj
Fl.
Ex.
ωj
Fl.
Muscle power
+
t1 t2 t3 t4 t5
–
ωj ωj ωj ωj
Mj Mj
Mj
Mj
Figure 6.6 Sequence of events during simple extension and flexion of the forearm. Muscle power shows two
positive bursts alternating with two negative bursts.
product of power and time is work, and it is measured in joules (1 J = 1 W s). If a muscle gen-
erates 100 W for 0.1 s, the mechanical work done is 10 J. This means that 10 J of mechanical
energy has been transferred from the muscle to the limb segments. As the example in
Figure 6.6 shows, power is continuously changing with time. Thus, the mechanical work done
must be calculated from the time integral of the power curve. The work done by a muscle during a
period t1 to t2 is:
t2
Wm = Pm dt J (6.2)
t1
In the example described, the work done from t1 to t2 is positive, from t2 to t3 it is negative,
from t3 to t4 it is positive again, and during t4 to t5 it is negative. If the forearm returns to the
starting position, the net mechanical work done by the muscles is zero, meaning that the time
integral of Pm from t1 to t5 is zero. It is therefore critical to know the exact times when Pm is
reversing polarities in order to calculate the total negative and the total positive work done during
the event.
6.0.8 Mechanical Work Done on an External Load

When any part of the body exerts a force on an adjacent segment or on an external body, it can
only do work if there is movement. In this case, work is defined as the product of the force acting
on a body and the displacement of the body in the direction of the applied force. The work, dW,
done when a force causes an infinitesimal displacement, ds, is:
dW = F ds (6.3)
ISTUDY
Or the work done when F acts over a distance S1 is:

S1
W= F ds = FS1 (6.4)
0
If the force is not constant (which is most often the case), then we have two variables that
change with time. Therefore, it is necessary to calculate the power as a function of time and inte-
grate the power curve with respect to time to yield the work done. Power is the rate of doing work,
or dW/dt.
dW ds
P= =F
dt dt (6.5)
=F V
where
P = instantaneous power, W
F = force, N
V = velocity, m/s.
Since both force and velocity are vectors, we must take the dot product, or the product of the
force and the component of the velocity that is in the same direction as the force. This will yield:
P = FV cos θ = F x V x + F y V y (6.6)
where
θ = angle between the force and velocity vectors in the plane defined by those vectors
Fx and Fy = forces in x and y directions
Vx and Vy = velocities in x and y directions.
For the purpose of this initial discussion, let us assume that the force and the velocity are
always in the same direction. Therefore, cos θ = 1 and:
P = FV W
t t (6.7)
W= P dt = FV dt J
0 0
Example 6.1. A baseball is thrown with a constant accelerating force of 100 N for a period of
180 ms. The mass of the baseball is 1.0 kg, and it starts from rest. Calculate the work done on the
baseball during the time of force application.
Solution
1 2
S1 = ut + at
2
u =0
a = F m = 100 1 0 = 100 m s2
1
S1 = × 100 0 18 2 = 1 62 m
2
S1
W= F ds = FS1 = 100 × 1 62 = 162 J
0
ISTUDY
100
80
Force (N)
60
40
20
0
20
16
Velocity (m/s)
12 t
8
V= ʃ0 a dt
4
0
1200
1000
P=F·V
Power (W)
800
600
400
200
0
150
125
ʃ0
t
100 W = P dt
Work (J)
75
50
25
0
0 40 80 120 160 200 240 280 320
Time (ms)
Figure 6.7 Forces, velocity, mechanical power, and work done on a baseball while being thrown. See the text for
details.
Example 6.2 A baseball of mass 1 kg is thrown with a force that varies with time, as indicated
in Figure 6.7. The velocity of the baseball in the direction of the force is also plotted on the
same time base and was calculated from the time integral of the acceleration curve (which
has the same numerical value as the force curve because the mass of the baseball is 1 kg).
Calculate the instantaneous power to the baseball and the total work done on the baseball during
the throwing period.
The peak power calculated here may be considered quite high, but it should be noted that this
peak has a short duration. The average power for the throwing period is less than 500 W. In real-
life situations, it is highly unlikely that the force will ever be constant; thus, instantaneous power
must always be calculated. When the baseball is caught, the force of the hand still acts against the
baseball, but the velocity is reversed. The force and velocity vectors are now in opposite direc-
tions. Thus, the power is negative and the work done is also negative, indicating that the baseball
is doing work on the body.
6.0.9 Mechanical Energy Transfer Between Segments

Each body segment exerts forces on its neighboring segments, and if there is a translational move-
ment of the joints, there is a mechanical energy transfer between segments. In other words, one
segment can do work on an adjacent segment by a force
ISTUDY
6.1 EFFICIENCY 123
Segm
Fj1
ent 1
θ1 Vj
Vj
θ2
Fj2
ent 2
Segm
Figure 6.8 Reaction forces and velocities at a joint center during dynamic activity. The dot product of the force
and velocity vectors is the mechanical power (rate of mechanical energy transfer) across the joint.
(Quanbury et al., 1975). This work is in addition to the muscular work described in Sections
6.0.4–6.0.7. Equations can be used to calculate the rate of energy transfer (i.e. power) across
the joint center. Consider the situation in Figure 6.8 at the joint between two adjacent segments.
Fj1, the reaction force of segment 2 on segment 1, acts at an angle θ1 from the velocity vector
Vj. The product of Fj1Vj cos θ1 is positive, indicating that energy is being transferred into
segment 1. Conversely, Fj2Vjcosθ2 is negative, denoting a rate of energy outflow from segment
2. Since Pj1 = − Pj2, the outflow from segment 2 equals the inflow to segment 1. In an n-joint sys-
tem, there will be n power flows, but the algebraic sum of all those power flows will be zero, rein-
forcing the fact that these flows are passive and, therefore, do not add to or subtract from the total
body energy.
This mechanism of energy transfer between adjacent segments is quite important in the con-
servation of energy of any movement because it is a passive process and does not require muscle
activity. In walking, this has been analyzed in detail (Winter and Robertson, 1978). At the end of
swing, for example, the swinging foot and leg lose much of their energy by transfer upward
through the thigh to the trunk, where it is conserved and converted to KE to accelerate the upper
body in the forward direction.
6.1 EFFICIENCY
The term efficiency is probably the most abused and misunderstood term in human movement
energetics. Confusion and error result from an improper definition of both the numerator and
the denominator of the efficiency equation (Gaesser and Brooks, 1975; Whipp and Wasserman,
1969). In the next section, four causes of inefficiency are discussed in detail, and these mechan-
isms must be recognized in whatever formula evolves. Overlaid on these four mechanisms are
two fundamental reasons for inefficiency: inefficiency in the conversion of metabolic energy to
mechanical energy, and neurological inefficiency in the control of that energy. Metabolic energy
is converted to mechanical energy at the tendon, and the metabolic efficiency depends on the
conditioning of each muscle, the metabolic (fatigue) state of muscle, the subject’s diet, and
any possible metabolic disorder. This conversion of energy would be called metabolic or muscle
efficiency and would be defined as follows:
Σ mechanical work done by all muscles

metabolic muscle efficiency = (6.8)
ISTUDY
Such an efficiency is impossible to calculate at this time because it is currently impossible to

calculate the work of each muscle (which would require force and velocity time histories of every
muscle involved in the movement) and to isolate the metabolic energy of those muscles. Thus, we
are forced to compromise and calculate an efficiency based on segmental work and to correct the
metabolic cost by subtracting estimates of overhead costs not associated with the actual mechan-
ical work involved. Thus, an efficiency would be defined as:
mechanical work internal + external

mechanical efficiency = (6.9)
metabolic cost − resting metabolic cost
The resting metabolic cost in bicycling, for example, could be the cost associated with sitting
still on the bicycle.
A further modification is work efficiency, which is defined as:
external mechanical work

work efficiency = (6.10)
metabolic cost − zero-work metabolic cost
The zero-work cost would be the cost measured with the cyclist freewheeling.
In all of the efficiency calculations described, there are varying amounts of positive and neg-
ative work. The metabolic cost of positive work exceeds that of equal levels of negative work.
However, negative work is not negligible in most activities. Level gait has equal amounts of pos-
itive and negative work. Running uphill has more positive work than negative work, and vice
versa for downhill locomotion. Thus, all of the efficiency calculations yield numbers that are
strongly influenced by the relative percentages of positive and negative work. An equation that
gets around this problem is:
metabolic cost positive work + metabolic cost negative work = metabolic cost
positive work negative work
or + = metabolic cost (6.11)
η+ η−
where η+ and η− are the positive and negative work efficiencies, respectively.
The interpretation of efficiency is faulty if it is assumed to be simply a measure of how well the
metabolic system converts biochemical energy into mechanical energy, rather than a measure of
how well the neural system is performing to control the conversion of that energy. An example
will demonstrate the anomaly that results. A normal healthy adult walks with 100 J mechanical
work per stride (half positive, half negative). The metabolic cost is 300 J per stride, and this yields
an efficiency of 33%. A neurologically disabled adult would do considerably more mechanical
work because of his or her jerky gait pattern, say 200 J per stride. Metabolically, the cost might
be 500 J per stride, which would give an efficiency of 40%. Obviously, the healthy adult is a more
efficient walker, but our efficiency calculation does not reflect that fact. Neurologically, the disa-
bled person is quite inefficient because he or she is not generating an effective and smooth neural
pattern. However, the disabled person is quite efficient in the actual conversion of metabolic energy
to mechanical energy (at the tendon), and that is all that is reflected in the higher efficiency score.
6.1.1 Causes of Inefficient Movement

It is often difficult for a therapist or coach to concentrate directly on efficiency. Rather, it is more
reasonable to focus on the individual causes of inefficiency and thereby automatically improve
the efficiency of the movement. The four major causes of mechanical inefficiency (Winter, 1978)
will now be described.
ISTUDY
6.1 EFFICIENCY 125
6.1.1.1 Cocontractions. Obviously, it is inefficient to have muscles cocontract because they

fight against each other without producing a net movement. Suppose that a certain movement can
be accomplished with a flexor moment of 30 N m. The most efficient way to do this is with flexor
activity only. However, the same movement can be achieved with 40 N m flexion and 10 N m
extension, or with 50 N m flexion and 20 N m extension. In the latter case, there is an unnecessary
20 N m moment in both the extensors and the flexors. Another way to look at this situation is that
the flexors are doing unnecessary positive work to overcome the negative work of the extensors.
Cocontractions occur in many pathologies, notably those involving the central nervous sys-
tem, as is the case in individuals with hemiplegic stroke and spastic cerebral palsy. Cocontrac-
tions also occur to a limited extent during normal movement when it is necessary to stabilize a
joint, especially if heavyweights are being lifted or at the ankle joint during walking or running.
At present, the measurement of unnecessary cocontractions is only possible by monitoring the
EMG activity of the antagonistic muscles. Without an exact EMG calibration versus tension
for each muscle, it is impossible to arrive at a quantitative measure of cocontraction. Falconer
and Winter (1985) presented a formula by which cocontraction can be quantified,
M antag
COCON = 2 × × 100 (6.12)
M agon + M antag
where Mantag and Magon are the moments of force of antagonists and agonists, respectively.
In the example reported, the antagonist activity results in an equal increase in agonist activity;
thus, the unnecessary activity must be twice that of the antagonist alone. If the antagonists created
a 20 N m extensor moment and the agonists generated a 50 N m flexor moment, %COCON
would be 40/70 × 100% = 57%. However, most movements involve continuously changing mus-
cle forces; thus, an agonist muscle at the beginning of the movement will likely reverse its role
and become an antagonist later on in the movement. Joint moments of force, as seen in many
graphs in Chapter 4, reverse their polarities many times; thus, a modification of Equation (6.12)
is needed to cope with these time-varying changes. Figure 6.9 demonstrates the profile of activity
of two antagonistic muscles during a given movement. The cross-hatching of muscles A and B
shows a common area of activity that indicates the cocontraction area. Thus, the percent cocon-
traction is defined as:
common area A & B
COCON = 2 × × 100 (6.13)
area A + area B
If EMG is the primary measure of relative tension in the muscle, we can suitably process the
raw EMG to yield a tension-related activation profile (Milner-Brown et al., 1973; Winter, 1976).
Cocontraction
Force-related activation
A B
% stride
Figure 6.9 Profiles of activity of two antagonist muscles, with cross-hatched area representing the cocontraction.
See the text for detailed discussion and analysis.
ISTUDY
The activity profiles of many common muscles look very much like those portrayed in Figure 6.9.
In this case, muscle A is the tibialis anterior, and muscle B is the soleus as seen over one walking
stride. Using Equation (6.13) for the profiles in Figure 6.5, %COCON was calculated to be 24%.
It needs to be noted that additional methods for calculating cocontraction exist. Equation (6.14)
has been used to quantify muscle cocontraction in clinical studies.
COCON = M lower M higher × M lower + M higher (6.14)
Mlower and Mhigher represent the EMG values from the lower and higher amplitude muscle,
respectively, at each point in time (Rudolph et al., 2000). In this equation, the magnitude of each
muscle is considered, as well as the magnitude of their relative cocontraction. From a clinical
perspective, this may be an important consideration because the overall compression force within
a joint is dependent not only on the relative cocontration between opposing muscles (ratio) but
also the magnitude of the individual muscle force. Interpreting the results from studies in which
cocontraction is calculated must be done so cautiously. The equation used to estimate cocontrac-
tion can dramatically alter the magnitude of the cocontraction value (Kellis et al., 2003).
6.1.1.2 Isometric Contractions Against Gravity. In normal dynamic movement, there is min-
imal muscle activity that can be attributed to holding limb segments against the forces of gravity.
This is because the momentum of the body and limb segments allows for a smooth interchange of
energy. However, in many pathologies, the movement is so slow that there are extended periods
of time when limb segments or the trunk are being held in near-isometric contractions. Spastic
cerebral palsy patients often crouch with their knee flexed, requiring excessive quadriceps activ-
ity to keep them from falling down. Other times, individuals with central nervous system disor-
ders may have poor coordination and hold their limb off the ground for a period of time prior to
swing-through. This would require muscle activity to maintain the limb in an unsupported posi-
tion against gravity.
Most movement assessments, be it clinical or research studies, do not quantify isometric work
against gravity because there is no movement involved. While quantification of EMG signals, or
the use of EMG-driven models, can be used to provide some measure of the efficiency of move-
ment, each muscle’s EMG would have to be calibrated against the extra metabolism required to
contract that muscle. At present, no there is no standard technique to separate the metabolic cost
of this inefficiency.
6.1.1.3 Generation of Energy at One Joint and Absorption at Another. The least known and
understood cause of inefficiency occurs when one muscle group at one joint does positive work at
the same time as negative work is being done at others. Such an occurrence is really an extension
of what occurs during a cocontraction (e.g. positive work being canceled out by negative work of
the antagonistic muscles). It is quite difficult to visualize when this happens. During normal walk-
ing, it occurs during double support when the energy increase of the push off leg takes place at the
same time as the weight-accepting leg absorbs energy. Figure 6.10 shows this point in gait: the
Negative
Positive work
work
Figure 6.10 Example of a point in time during gait that positive work by the push off muscles can be canceled by
negative work of the weight-accepting muscles of the contralateral leg.
ISTUDY
6.1 EFFICIENCY 127
left leg’s push off (positive work) is due primarily to plantarflexors; the right leg’s energy absorp-
tion (negative work) takes place in the quadriceps and tibialis anterior. There is no doubt that the
instability of pathological gait is a major cause of this type of inefficient muscle activity. The only
way to analyze such inefficiencies is to calculate the muscle power at each joint separately and to
quantify the overlap of simultaneous phases of positive and negative work.
In spite of the inefficiencies inferred by such events, it must be remembered that many com-
plex movements such as walking or running require that several functional tasks be performed at
the same time. The example in Figure 6.10 illustrates such a situation: the plantarflexors are com-
pleting push off while the contralateral muscles are involved in weight acceptance. Both these
events are essential to a safe walking pattern.
6.1.1.4 Jerky Movements. Efficient energy exchanges are characterized by smooth-looking

movements. A ballet dancer and a high jumper execute smooth movements for different reasons,
one for artistic purposes, the other for efficient performance. Energy added to the body by pos-
itive work at one point in time is conserved, and very little of this energy is lost by muscles doing
negative work. The jerky gait of a child with cerebral palsy is quite the opposite. Energy added at
one time is removed a fraction of a second later. The movement has a steady succession of stops
and starts, and each of these bursts of positive and negative work has a metabolic cost. The energy
cost from jerky movements can be assessed in two ways: by work analysis based on a segment-
by-segment energy analysis or by a joint-by-joint power analysis. Both of these techniques are
described later.
6.1.2 Summary of Energy Flows

It is valuable to summarize the flows of energy from the metabolic level through to an external
load. Figure 6.11 depicts this process schematically. Metabolic energy cannot be measured
directly but can be calculated indirectly from the amount of O2 required or by the CO2 expired.
The details of these calculations and their interpretation are the subject of many textbooks and are
beyond the scope of this book.
At the basal level (resting, lying down), the muscles are relaxed, but they still require metabolic
energy to keep them alive. The measure of this energy level is called maintenance heat. Then as a
muscle contracts, it requires energy, which shows up as additional heat called activation heat. It
has been shown to be associated with the rate of build-up of tension within the muscle and is
accompanied by an internal shortening of the muscle contractile elements. Stable heat is the heat
that measures the energy required to maintain tension within accounted for the muscle. Labile
heat is a third type of heat seen in isometric contractions and is the heat not generated by either
tension or rate of tension generation. The final type of heat loss is shortening heat, which is asso-
ciated with the actual shortening of the muscle under load. Students are referred to the excellent
review by Hill (1960).
O2 uptake Maintenance Isometric work

heat against gravity
Mechanical energy Increase in External

Metabolic
segment
energy Muscle tension work
energies
Heats of contraction Loss due to cocontraction

CO2 expired Activation, labile or absorption by muscles
stable, shortening at another joint
Figure 6.11 Flow of energy from metabolic level of external mechanical work. Energy is lost as heat associated
with the contractile process or by inefficiencies after energy has been
ISTUDY
Finally, at the tendon, we see the energy in mechanical form. The muscle tension can contrib-
ute to four types of mechanical loads. It can be involved with a cocontraction, isometric “work”
against gravity, or a simultaneous absorption by another muscle, or it can cause a net change in
the body energy. In the latter case, if positive work is done, the net body energy will increase; if
negative work is done, there will be a decrease in total body energy. Finally, if the body exerts
forces on an external body, some of the energy may be transferred as the body performs exter-
nal work.
6.2 FORMS OF ENERGY STORAGE
1. Potential Energy. PE is the energy due to gravity and, therefore, increases with the height
of the body above ground or above some other suitable reference datum,
PE = mgh J (6.15)
where
m = mass, kg
g = gravitational acceleration, equal to 9.8 m/s2
h = height of center of mass, m.
With h = 0, the PE decreases to zero. However, the ground reference datum should be carefully
chosen to fit the problem in question. Normally, it is considered to be the lowest point that the
body takes during the given movement. For a diver, it could be the water level; for a person walk-
ing, it would be the lowest point in the pathway.
2. Kinetic Energy. There are two forms of KE, that due to translational velocity and that due
to rotational velocity,
translational KE = 1 2mv2 (6.16)
where v = velocity of center of mass, m/s.
rotational KE = 1 2Iω2 (6.17)
where
I = rotational moment of inertia, kg m2
ω = rotational velocity of segment, rad/s.
Note that these two energies increase as the velocity is squared. The polarity of direction of the
velocity is unimportant because velocity squared is always positive. The lowest level of KE is
therefore zero when a body is at rest.
3. Total Energy and Exchange within a Segment. As mentioned previously, the energy of a
body exists in three forms so that the total energy of a body is:
E s = PE + translational KE + rotational KE
1 1 (6.18)
= mgh + mv2 + Iω2 J
2 2
It is possible for a body to exchange energy within itself and still maintain a constant total
energy.
ISTUDY
6.2 FORMS OF ENERGY STORAGE 129
Example 6.3. Suppose that the baseball in the Example is thrown vertically. Calculate the
potential and kinetic energies at the time of release, at maximum height, and when it reaches
the ground. Assume that it is released at a height of 2 m above the ground and that the vertical
accelerating force of 100 N is in excess of gravitational force. At release,
a = 100 m s2 as calculated previously

t1
v = a dt = at 1 = 100t 1
0
t 1 = 180 ms
v = 18 m s
1 2 1
translational KE = mv = × 1 × 182 = 162 J
2 2
Note that this 162 J is equal to the work done on the baseball prior to release.
total energy = PE t 1 + translational KE t 1

= 19 6 + 162 = 181 6 J
If we ignore air resistance, the total energy remains constant during the flight of the baseball,
such that at t2, when the maximum height is reached, all the energy is PE and KE = 0. Therefore,
PE(t2) = 181.6 J. This means that the baseball reaches a height such that mgh2 = 181.6 J, and:
181 6
h2 = = 18 5 m
10×98
At t3, the baseball strikes the ground, and h = 0. Thus, PE(t3) = 0 and KE(t3) = 181.6 J. This
means that the velocity of the baseball is such that 12 mv2 = 181 6 J, or:
v = 19 1 m s
This velocity is slightly higher than the release velocity of 18 m/s because the ball was released
from 2 m above ground level.
6.2.1 Energy of a Body Segment and Exchanges of Energy Within the Segment
Most body segments contain all three energies in various combinations at any point in time during
a given movement. A diver at the top of a dive has considerable PE, and during the dive, converts
it to KE. Similarly, a boomerang when released has rotational and translational KE, and at peak
height, some of the translational KE has been converted to PE. At the end of its travel, the boo-
merang will have regained most of its translational KE.
In a multisegment system, such as the human body, the exchange of energy can be consider-
ably more complex. There can be exchanges within a segment or between adjacent segments.
A good example of energy exchange within a segment is during normal gait. The upper part
of the body (head, arms, and trunk [HAT]) has two peaks of PE each stride – during the midstance
of each leg. At this time, HAT has slowed its forward velocity to a minimum. Then, as the body
falls forward to the double-support position, HAT picks up velocity at the expense of a loss in
ISTUDY
1.29
Vertical Ht. (m)

1.28
1.27
1.26
1.25
1.24
Horiz. velocity (m/s)

1.7
1.6
1.5
1.4
1.3
1.2
Time
Figure 6.12 Plot of vertical displacement and horizontal velocity of HAT shows evidence of energy exchange
within the upper part of the body during gait.
Simple pendulum
Total energy
Potential Kinetic
energy energy
Time
Figure 6.13 Exchange of kinetic and potential energies in a swinging frictionless pendulum. Total energy of the
system is constant, indicating that no energy is being added or lost. (With permission of University of
Toronto Press.)
height. Evidence of energy exchange should be seen from a plot of the horizontal velocity and the
vertical displacement of the center of gravity of HAT (see Figure 6.12). The PE, which varies
with height, changes roughly as a sinusoidal wave, with a minimum during double support
and reaching a maximum during midstance. The forward velocity is almost completely out of
phase, with peaks approximately during double support and minima during midstance.
Exchanges of energy within a segment are characterized by opposite changes of the potential
and KE components. Figure 6.13 shows what would happen if a perfect exchange took place, as
in a swinging frictionless pendulum. The total energy would remain constant over time in the
presence of large changes of potential and KE.
Consider the other extreme, in which no energy exchange takes place. Such a situation would
be characterized by totally in-phase energy components, not necessarily of equal magnitude, as
depicted in Figure 6.14.
6.2.1.1 Approximate Formula for Energy Exchanges Within a Segment. An approximation

of energy exchange can be calculated if we know the peak-to-peak change in each of the energy
components over a period of time:
E ex = ΔEp + ΔEkt + ΔE kr − ΔE s (6.19)
If there is no exchange, ΔEp + ΔEkt + ΔEkr = ΔEs. If there is 100% exchange, ΔE = 0.
ISTUDY
6.2 FORMS OF ENERGY STORAGE 131
Total
Energy
Potential
Translational
kinetic
Rotational
kinetic
Time
Figure 6.14 Energy patterns for a segment in which no exchanges are taking place. All energy components are
perfectly in phase.
Example 6.4. A visual scan of the energies of the leg segment during walking yields the fol-
lowing maximum and minimum energies on the stride period:
E s max = 29 30 J, E s min = 13 14 J,
E p max = 15 18 J, E p min = 13 02 J,
E kt max = 13 63 J, Ekt min = 0 09 J,
E kr oxmax = 0 95 J, E kr min = 0 J
Thus,
ΔEs = 29 30 − 13 14 = 16 16 J,
ΔEp = 15 18 − 13 02 = 2 16 J,
ΔEkt = 13 63 − 0 09 = 13 54 J,
ΔEkr = 0 95 − 0 = 0 95 J
Since ΔEp + ΔEkt + ΔE kr = 16 65 J, it can be said that 16.65 − 16.16 = 0.49 J were exchanged
during the stride. Thus, the leg is a highly nonconservative system.
6.2.1.2 Exact Formula for Energy Exchange Within Segments. The example just discussed
illustrates a simple situation in which only one minimum and maximum occur over the period of
interest. If individual energy components have several maxima and minima, we must calculate
the sum of the absolute energy changes over the time period. The work, Ws, done on and by a
segment during N sample periods is:
N
Ws = ΔE s J (6.20)
i=1
Assuming that there are no energy exchanges between any of the three components (Norman
et al., 1976), the work done by the segment during the N sample periods is:
N
Ws = ΔE p + ΔE kt + ΔEkr J (6.21)
i=1
ISTUDY
Therefore, the energy Wc conserved within the segment during the time is:
Wc = Ws − Ws J (6.22)
The percentage energy conservation, Cs, during the time of this event is:
Wc
Cs = × 100 (6.23)
Ws
If W s − W s, all three energy components are in phase (they have exactly the same shape and
have their minima and maxima at the same time), and there is no energy conservation. Con-
versely, as demonstrated by an ideal pendulum, if Ws = 0, then 100% of the energy is being
conserved.
6.2.2 Total Energy of a Multisegment System

As we proceed with the calculation of the total energy of the body, we merely sum the energies of
each of the body segments at each point in time (Bresler and Berry, 1951; Ralston and Lukin,
1969; Winter et al., 1976). Thus, the total body energy Eb at a given time is:
B
Eb = Esi J (6.24)
i=1
where
Esi = total energy of ith segment at that point in time
B = number of body segments.
The individual segment energies continuously change with time, so it is not surprising that the
sum of these energies will also change with time. However, the interpretation of changes in Eb
must be done with caution when one considers the potential for transfer of energy between seg-
ments and the number of possible generators and absorbers of energy at each joint. For example,
transfers of energy between segments (see Section 6.0.9) will not result in either an increase or a
loss in total body energy. However, in other than the simplest movements, there may be several
simultaneous concentric and eccentric contractions. Thus, over a period of time, two muscle
groups may generate 30 J, while one other muscle group may absorb 20 J. The net change in body
energy over that time would be an increase of 10 J. Only through a detailed analysis of mechan-
ical power at the joints (see Section 6.3.1.4) are we able to assess the extent of such a cancelation.
Such events are obviously inefficient but are necessary to accomplish many desired movement
patterns.
Consider now a simple muscular system that can be represented by a pendulum mass with a
pair of antagonistic muscle groups, m1 and m2, crossing a simple hinge joint. Figure 6.15 shows
such an arrangement along with a time history of the total energy of the system. At t1, the segment
is rotating counterclockwise at ω1 rad/s. No muscle activity occurs until t2 when m2 contracts.
Between t1, and t2, the normal pendulum energy exchange takes place and the total energy
remains constant. However, between t2 and t3, muscle m2 causes an increase in both kinetic
and potential energies of the segment. The muscle moment has been in the same direction as
the direction of rotation, so positive work has been done by the muscle on the limb segment,
and the total energy of the limb has increased. Between t3 and t4, both muscles are inactive,
and the total energy remains at the higher but constant
ISTUDY
6.3 CALCULATION OF INTERNAL AND EXTERNAL WORK 133
m1 m1 m1 m1 m1
m2
m2 m2 m2 m2
Energy
t1 t2 t3 t4 t5
Figure 6.15 Pendulum system with muscles. When positive work is done, the total energy increases; when
negative work is done, the total energy decreases.
down the segment. Energy is lost by the segment and is absorbed by m1. This is negative work
being done by the muscle because it is lengthening during its contraction. Thus, at t5 the segment
has a lower total energy than at t4.
The following major conclusions can be drawn from this example.
1. When muscles do positive work, they increase the total body energy.
2. When muscles do negative work, they decrease the total body energy.
3. During cyclical activity, such as constant velocity level running, the net energy change per
stride equals zero (neglecting the small air friction and shoe friction losses). Thus, the inter-
nal positive work done per stride should equal the internal negative work done per stride.
6.3 CALCULATION OF INTERNAL AND EXTERNAL WORK
Internal and external work has been calculated in a multitude of ways by different researchers.
Some assume that the energy changes in the body center of mass yield the total internal work done
by all the muscles. Others look only at the “vertical work” resulting from PE increases in the body
center of gravity, while others (especially in the exercise physiology area) even ignore internal
work completely. It is, therefore, important to look at all possible sources of muscle activity that
have a metabolic cost and keep those readily available on a “checklist” to see how complete any
analysis really is. This same list serves a useful purpose in focusing our attention on possible
causes of inefficient movement (see Section 6.1.1).
6.3.1 Internal Work Calculation

The various techniques for calculating the internal work have undergone a general improvement
over the years. The vast majority of the research has been done in the area of human gait, and
because gait is a complex movement, it will serve well as an example of the dos and don’ts.
6.3.1.1 Energy Increases in Segments. A number of early researchers attempted a calcula-

tion of work based on increases in potential or kinetic energies of the body or of individual seg-
ments. Fenn (1929), in his accounting of the flow of energy from metabolic to mechanical,
calculated the kinetic and potential energies of each major segment of a sprinter. He then summed
the increases in each of these segment energies over the stride period to yield the net mechanical
work. Unfortunately, Fenn’s calculations ignored two
ISTUDY
energy exchanges within segments and passive transfers between segments. Thus, his mechanical
work calculations were predictably high: the average power of his sprinters was computed to be
three horsepower. Conversely, Saunders et al. (1953), Cotes and Meade (1960), and Liberson
(1965) calculated the “vertical work” of the trunk as representing the total work done by the body.
These calculations ignored the major energy exchange that takes place within the HAT and also
the major work done by the lower limbs.
6.3.1.2 Center-of-Mass Approach. Cavagna and Margaria (1996) proposed a technique to

calculate work based on potential and kinetic energies of the body’s center of mass. Their data
were based on force platform records during walking and running, from which the translational
kinetic and potential energies were calculated. Such a model makes the erroneous assumption that
the body’s center of mass reflects the energy changes in all segments. The body’s center of mass
is a vector sum of all segment mass-acceleration products, and, as such, opposite-polarity accel-
erations will cancel out. But energies are scalars, not vectors, and, therefore, the reciprocal move-
ments that dominate walking and running will be largely canceled. Thus, simultaneous increases
and decreases in oppositely moving segments will go unnoticed. Also, Cavagna’s technique is
tied to force platform data, and nothing is known about the body’s center of gravity during non-
weight-bearing phases of running. Thus, this technique has underestimation errors and limita-
tions that have been documented (Winter, 1979). Also, the center-of-mass approach does not
account for the energy losses from the simultaneous generation and absorption of energy at dif-
ferent joints.
6.3.1.3 Sum of Segment Energies. A major improvement on the previous techniques was
made by Ralston and Lukin (1969) and Winter et al. (1976). Using displacement transducers
and television (TV) imaging techniques, the kinetic and potential energies of the major segments
were calculated. A sum of the energy components within each segment recognized the conser-
vation of energy within each segment (see Section 6.2.1) and a second summation across all
segments recognized energy transfers between adjacent segments (see Section 6.0.9). The total
bodywork is calculated (Winter, 1979) to be:
N
Wb = ΔEb J (6.25)
i=1
However, this calculation underestimates the simultaneous energy generation and absorption
at different joints. Thus, Wb will reflect a low estimate of the positive and negative work done by
the human motor system. Williams and Cavanagh (1983) made empirical estimates to correct for
these underestimates in running.
6.3.1.4 Joint Power and Work. In Sections 6.0.6 and 6.0.7, techniques for the calculation of
the positive and negative work at each joint were presented. Using the time integral of the
power curve (Equation 6.6), we are able to get at the “sources” and “sinks” of all the mechanical
energy. Figure 6.16 is an example to show the work phases at the knee during slow running.
The power bursts are labeled K1, , K5, and the energy generation/absorption resulting from
the time integral of each phase is shown (Winter, 1983). In this runner, it is evident that the
energies absorbed early in stance (53 J) by the knee extensors and by the knee flexors in later
swing (24 J) dominate the profile; only 31 J are generated by the knee extensors in middle and
late stance.
It should be noted that this technique automatically calculates any external work that is done.
The external power will be reflected in increased joint moments, which, when multiplied by the
joint angular velocity, will show an increased power equal to that done externally.
ISTUDY
6.3 CALCULATION OF INTERNAL AND EXTERNAL WORK 135
Knee
100° Flexing Flexing Extending
Extending WN20P
80°
θk 60°
40°
20°
HC TO HC
0
0.2 0.4 0.6 0.8
Time (s)
210
180
150
Moment of force (N·m)
120
Ext.
90
60
30
0
0.2 0.4 0.6 0.8
fl.
–30
200 K2
K5
31 J
5J
Muscle power (w)
0 –11 J –24 J
0.2 0.4 0.6 0.8
–53 J K3
–200 K4
–400 K1
Figure 6.16 Plots of knee angle, moment of force, and power in a slow runner. Five power phases are evident: K1,
energy absorption by knee extensors; K2, positive work as extensors shorten under tension; K3, deceleration of
backward rotation of leg and foot as thigh drives forward during late stance and early swing; K4, deceleration of
swinging leg and foot by knee flexors prior to heel contact; and K5, small positive burst to flex the leg slightly
and slow down its forward motion to near-zero velocity prior to heel contact. (From Winter (1983)/With
permission of Elsevier.)
6.3.1.5 Muscle Power and Work. Even with the detailed analysis described in the previous
section, we have underestimated the work done by cocontracting muscles. Joint power, as
calculated, is the product of the joint moment of force Mj and the angular velocity ωj. Mj is
the net moment resulting from all agonist and antagonist activity, and therefore, cannot account
for simultaneous generation by one muscle group and absorption by the antagonist group, or
ISTUDY
vice versa. For example, if Mj = 40 N m and ωj = 3 rad/s, the joint power would be calculated
to be 120 W. However, if there were a cocontraction, the antagonists might be producing
a resisting moment of 10 N m. Thus, in this case, the agonists would be generating energy
at the rate of 50 × 3 = 150 W while the antagonists would be absorbing energy at a rate of
10 × 3 = 30 W.
Thus, the net power and work calculations as described in Section 6.3.1.4 will underestimate
both the positive and the negative work done by the muscle groups at each joint. To date, there has
been very limited progress to calculate the power and work associated with each muscle’s action.
The major problem is to partition the contribution of each muscle to the net moment, and this
issue has been addressed in Section 5.3.1. However, if the muscle force Fm and the muscle veloc-
ity Vm were known, the muscle work Wm would be calculated as:
t2
Wm = F m V m dt (6.26)
t1
Morrison (1970) analyzed the power and work in four muscles in normal walking, and some
later work (Yack, 1986) analyzed the muscles forces and powers in the three major biarticulate
muscle groups during walking.
6.3.1.6 Summary of Work Calculation Techniques. Table 6.1 summarizes the various
approaches described over the past few decades and the different energy components that are
not accounted for by each technique.
6.3.2 External Work Calculation

It was noted in Sections 6.3.1.4 and 6.3.1.5 that the work calculations done using Equations (6.6)
and (6.26) automatically take into account all work done by the muscles independent of whether
that work was internal or external. There is no way to partition the external work except by taking
measurements at the interface between human and external load. A cyclist, for example, would
require a force transducer on both pedals plus a measure of the velocity of the pedal. Similarly, to
analyze a person lifting or lowering a load would need a force transducer between the hands and
the load, or an imaging record of the load and the body (from which an inverse solution would
calculate the reaction forces and velocity). The external work We is calculated as:
t2
We = F r V c dt (6.27)
t1
where
F r = reaction force vector, N
V c = velocity of contact point, m/s
t1, t2 = times of beginning and end of each power phase.
6.4 POWER BALANCES AT JOINTS AND WITHIN SEGMENTS
In Section 6.1.2, examples were presented to demonstrate the law of conservation of mechanical
energy within a segment. Also, in Section 6.0.6, muscle mechanical power was introduced, and in
Section 6.0.9, the concept of passive energy transfers across joints was noted. We can now look at
one other aspect of muscle energetics that is necessary before we can achieve a complete power
ISTUDY
6.4 POWER BALANCES AT JOINTS AND WITHIN SEGMENTS 137
TABLE 6.1 Techniques to Calculate Internal Work in Movement

Technique Work Components Not Accounted for by Technique
Increase PE or Energy Simultaneous Simultaneous Cocontractions Work
KE (Fenn, exchange increases or generation and against g
1929; within decreases in absorption at
Saunders segments and reciprocally different joints
et al., 1953; transfers moving
Liberson, between segments
1965) segments
Center of mass Simultaneous Simultaneous Cocontractions Work
(Cavagna and increases or generation and against g
Margaria, decreases in absorption at
1996) reciprocally different joints
moving
segments
Σ Segment Simultaneous Cocontractions Work
energies generation and against g
(Winter, 1979) absorption at
different joints
Joint power* Cocontractions Work
Mjωj dt against g
(Winter, 1983)
Muscle power* Work
FmVm dt against g
(Yack, 1986)
* Also accounts for external work if it is present.
balance segment by segment: the fact that active muscles can transfer energy from segment to
segment in addition to their normal role of generation and absorption of energy.
6.4.1 Energy Transfer via Muscles

Muscles can function to transfer energy from one segment to the other if the two segments are
rotating in the same direction. In Figure 6.17, we have two segments rotating in the same direc-
tion but with different angular velocities. The product of Mω2 is positive (both M and ω2 have
the same polarity), and this means that energy is flowing into segment 2 from the muscles
responsible for moment M. The reverse is true as far as segment 1 is concerned, Mω1 is neg-
ative, showing that energy is leaving that segment and entering the muscle. If ω1 = ω2 (i.e. an
isometric contraction), the same energy rate occurs and a transfer of energy from segment 1 to
segment 2 via the isometrically acting muscles. If ω1 > ω2, the muscles are lengthening, and
thus, absorption plus a transfer takes place, while if ω1 < ω2, the muscles are shortening and a
generation as well as a transfer occur. Figure 6.18, from Robertson and Winter (1980), sum-
marizes all possible power functions that can occur at a given joint, and if we do not account for
these energy transfers through the muscles, we will not be able to account for the total power
balance within each segment. Thus, we must modify Equation (6.5) to include the angular velo-
cities of the adjacent segments in order to partition the transfer component. Therefore, ωj is
replaced by (ω1 − ω2),
Pm = M j ω1 − ω2 W (6.28)
ISTUDY
Seg
me
nt
ω1
1
M
2
ω2
Segment
Figure 6.17 Energy transfer between segments occurs when both segments are rotating in the same direction and
when there is a net moment of force acting across the joint. See the text for a detailed discussion.
ω1 ω1
M M
ω2 ω2
ω1
M M
ω2 ω2
ω1
M M
ω2 ω2
ω1
M M
ω2
Figure 6.18 Power generation, transfer, and absorption functions. (From Roberston and Winter (1980)/With
permission of Elsevier.)
Thus, if ω1 and ω2 have the same polarity, the rate of transfer will be the lesser of the two power
components. Examples are presented in Section 6.4.2 to demonstrate the calculation and to rein-
force the sign convention used.
6.4.2 Power Balance Within Segments

Energy can enter or leave a segment at muscles and across joints at the proximal and distal ends.
Passive transfer across the joint (Equation 6.9) and active
ISTUDY
6.4 POWER BALANCES AT JOINTS AND WITHIN SEGMENTS 139
(a) (b)
Fyp
Pjp = FxpVxp + FypVyp
Vyp
Pmp = Mp ωs
Vxp
Fxp
Mp dEs
= Pjp + Pmp + Pjd + Pmd
dt
ωs Fyd
Vyd
Vxd Pmd = Md ωs
Fxd
Md
Pjd = FxdVxd + FydVyd
Figure 6.19 (a) Biomechanical variables describing the instantaneous state of a given segment in which passive
energy transfers may occur at the proximal and distal joint centers and active transfers through the muscles at the
proximal and distal ends. (b) Power balance as calculated using the variables shown in (a). The passive power flow at
the proximal end Pjp, and the distal end Pjd, combined with the active (muscle) power at the proximal end Pmp and the
distal end Pmd, must equal the rate of change of energy in the segment dEs/dt.
(Equation 6.28) must be calculated. Consider Figure 6.19a as the state of a given segment at
any given point in time. The reaction forces and the velocities at the joint centers at the proximal
and distal ends are shown plus the moments of force acting at the proximal and distal ends along
with the segment angular velocity. The total energy of the segment Es as calculated by Equa-
tion (6.18) must also be known. Figure 6.19b is the power balance for that segment, the arrows
showing the directions where the powers are positive (energy entering the segment across the
joint or through the tendons of the dominant muscles). If the force–velocity or moment–ω product
turns out to be negative, this means that energy flow is leaving the segment. According to the law
of conservation of energy, the rate of change of energy of the segment should equal the four
power terms,
dEs
= Pjp + Pmp + Pjd + Pmd (6.29)
dt
(a) Biomechanical variables describing the instantaneous state of a given segment in which
passive energy transfers may occur at the proximal and distal joint centers and active trans-
fers through the muscles at the proximal and distal ends.
(b) Power balance as calculated using the variables shown in (a). The passive power flow at
the proximal end Pjp, and the distal end Pjd, combined with the active (muscle) power at
the proximal end Pmp and the distal end Pmd, must equal the rate of change of energy of the
segment dEs/dt.
A sample calculation for two adjacent segments is necessary to demonstrate the use of such
power balances and also to demonstrate the importance of passive transfers across joints and
across muscles as major mechanisms in the energetics of human movement.
Example 6.5. Carry out a power balance for the leg and thigh segments for frame 5, that is,
deduce the dynamics of energy flow for each segment separately and determine the power
dynamics of the knee muscles (generation, absorption,
ISTUDY
Table A.2a, hip velocities,
V xh = 1 36 m s V yh = 0 27 m s
Table A.2b, knee velocities,
V xk = 2 61 m s V yk = 0 37 m s
Table A.2c, ankle velocities,

V xa = 3 02 m s V ya = 0 07 m s
Table A.3b, leg angular velocity,

ωlg = 1 24 rad s
Table A.3c, thigh angular velocity,
ωth = 3 98 rad s
Table A.5a, leg segment reaction forces and moments,
F xk = 15 1 N, F yk = 14 6 N, F xa = − 12 3 N, F ya = 5 5 N,
Ma = − 1 1 N m Mk = 5 8 N m
Table A.5b, thigh segment reaction forces and moments,
F xk = − 15 1 N, F yk = − 14 6 N, F xh = − 9 4 N, F yk = 102 8 N,
M k = − 5 8 N m, M h = 8 5 N m
Table A.6, leg energy,
Elg frame6 = 20 5 J, E lg frame4 = 20 0 J
Table A.6, thigh energy,
Eth frame6 = 47 4 J, E th frame4 = 47 9 J
1. Leg Power Balance.
Σ powers = F xk V xk + F yk V yk + M k ωlg + F xa V xa + F ya V ya + M a ωlg

= 15 1 × 2 61 + 14 6 × 0 37 + 5 8 × 1 24 − 12 3
× 3 02 + 5 5 × 0 7 − 1 1 × 1 24
= 44 81 + 7 19 − 33 3 − 1 36
= 17 34 W
ΔElg 20 5 − 20 0
= = 17 5 W
Δt 0 0286
balance = 17 5 − 17 34 = 0 16 W
ISTUDY
6.5 PROBLEMS BASED ON KINETIC AND KINEMATIC DATA 141
14.97 33.83
ΔEth
= –17.5
Δt
23.08
44.81 15.89
7.19
ΔElg
= –17.5
Δt
1.36
33.3
Figure 6.20 Summary of thigh and leg power flows as calculated in Example 6.5. There is power transfer from
the thigh to the leg at a rate of 7.19 W through the quadriceps muscles plus passive flow across the knee of
44.81 W.
2. Thigh Power Balance.
Σ powers = F xh V xh + F yh V yh + M h ωth + F xk V xk + F yk V yk + M k ωth

= − 9 4 × 1 36 + 102 8 × 0 27 + 8 5 × 3 98
− 15 1 × 2 61 − 14 6 × 0 37 − 5 8 × 3 98
= 14 97 + 33 83 − 44 81 − 23 08
= − 19 09 W
ΔE th 47 4 − 47 9
= = − 17 5 W
Δt 0 0286
balance = − 17 5 − − 19 09 = 1 59 W
3. Summary of Power Flows. The power flows are summarized in Figure 6.20 as follows:
23.08 W leaves the thigh into the knee extensors, and 7.19 W enters the leg from the same
extensors. Thus, the knee extensors are actively transferring 7.19 W from the thigh to the
leg and are simultaneously absorbing 15.89 W.
6.5 PROBLEMS BASED ON KINETIC AND KINEMATIC DATA
1. (a) Calculate the potential, translational, and rotational kinetic energies of the leg segment
for frame 20 using appropriate kinematic data, and check your answer with Table A.6
in Appendix A.
(b) Repeat Problem 1a for the thigh segment for frame 70.
ISTUDY
2. (a) Plot (every second frame) the three components of energy plus the total energy of the
leg over the stride period and discuss whether this segment conserves or does not con-
serve energy over the stride period (frames 28–97).
(b) Repeat Problem 2a for the thigh segment.
(c) Repeat Problem 2a for the HAT segment. Using Equation (6.19), calculate the approx-
imate percentage of energy conservation in the HAT segment over the stride period.
Compare this percentage with that calculated using the exact Equations (6.20)–(6.23).
3. (a) Assuming symmetrical gait, calculate the total energy of the body for frame 28. [Hint:
For a stride period of 68 frames, data for the left side of the body can be estimated using
right-side data half a stride (34 frames) later.]
(b) Scan the total energy of all segments, and note the energy changes over the stride
period in the lower limb compared with that of the HAT. From your observations,
deduce whether the movement of the lower limbs or that of HAT makes the major
demands on the metabolic system.
4. (a) Using segment angular velocity data in Table A.7 in the appendix plus appropriate data
from other tables, calculate the power generation or absorption of the muscles at the
following joints. Identify in each case the muscle groups involved that are responsible
for the generation or absorption. Check your numerical answers with Table A.7.
(i) Ankle joint for frame 30.
(ii) Ankle joint for frame 50.
(iii) Ankle joint for frame 65.
(iv) Knee joint for frame 35.
(v) Knee joint for frame 40.
(vi) Knee joint for frame 65.
(vii) Knee joint for frame 20.
(viii) Hip joint for frame 50.
(ix) Hip joint for frame 70.
(x) Hip joint for frame 4.
(b) (i) Scan the listings for muscle power in Table A.7 and identify where the major
energy generation occurs during walking. When in the gait cycle does this occur
and by what muscles?
(ii) Do the knee extensors generate any significant energy during walking? If so,
when during the walking cycle?
(iii) What hip muscle group generates energy to assist the swinging of the lower limb?
When is this energy generated?
5. Using Equation (6.10) (see Figure 6.8), calculate the passive rate of energy transfer across
the following joints, and check your answers with Table A.7. From what segment to what
segment is the energy flowing?
(i) Ankle for frame 20.
(ii) Ankle for frame 33.
(iii) Ankle for frame 65.
(iv) Knee for frame 2.
(v) Knee for frame 20.
(vi) Knee for frame 65.
(vii) Hip for frame 2.
(viii) Hip for frame 20.
(ix) Hip for frame 67.
ISTUDY
6.6 REFERENCES 143
6. (a) Using equations in Figure 6.19b, carry out a power balance for the foot segment for
frame 20.
(b) Repeat Problem 6(a) for the leg segment for frame 20.
(c) Repeat Problem 6(a) for the leg segment for frame 65.
(d) Repeat Problem 6(a) for the thigh segment for frame 63.
7. Muscles can transfer energy between adjacent segments when they are rotating in the same
direction in space. Calculate the power transfer between the following segments, and indi-
cate the direction of energy flow. Compare your answers with those listed in Table A.7.
(a) Leg/foot for frame 60.
(b) Thigh/leg for frame 7.
(c) Thigh/leg for frame 35.
6.6 REFERENCES
Abbot, B. C., B. Bigland, and J. M. Ritchie. “The Physiological Cost of Negative Work,” J. Physiol. 117:
380–390, 1952.
Bresler, B. and F. Berry. “Energy Levels during Normal Level Walking,” Report of Prosthetic Devices Res.
Proj., University of California, Berkeley, 1951.
Cappozzo, A., F. Figura, and M. Marchetti. “The Interplay of Muscular and External Forces in Human
Ambulation,” J. Biomech. 9: 35–43, 1976.
Cavagna, G. A. and R. Margaria. “Mechanics of Walking,” J. Appl. Physiol. 21: 271–278, 1996.
Cotes, J. and F. Meade. “The Energy Expenditure and Mechanical Energy Demand in Walking,” Ergonomics
3: 97–119, 1960.
Elftman, H. “Forces and Energy Changes in the Leg during Walking,” Am. J. Physiol. 125: 339–356, 1939.
Falconer, K. and D. A. Winter. “Quantitative Assessment of Cocontraction at the Ankle Joint in Walking,”
Electromyogr. Clin. Neurophysiol. 25: 135–148, 1985.
Fenn, W. D. “Frictional and Kinetic Factors in the Work of Sprint Running,” Am. J. Physiol. 92: 583–
611, 1929.
Gaesser, G. A. and G. A. Brooks. “Muscular Efficiency during Steady Rate Exercise: Effects of Speed and
Work Rate,” J. Appl. Physiol. 38: 1132–1139, 1975.
Hill, A. V. “Production and Absorption of Work by Muscle,” Science 131: 897–903, 1960.
Kellis, E., F. Arabatzi, and C. Papadopoulos. “Muscle Co-Activation around the Knee in Drop Jumping Using
the Co-Contraction Index,” J. Electromyogr. Kinesiol. 13(3): 229–238, 2003.
Liberson, W. T. “Biomechanics of Gait: A Method of Study,” Arch. Phys. Med. Rehab. 46: 37–48, 1965.
Milner-Brown, H., R. Stein, and R. Yemm. “The Contractile Properties of Human Motor Units during Vol-
untary Isometric Contractions,” J. Physiol. 228: 288–306, 1973.
Morrison, J. B. “Mechanics of Muscle Function in Locomotion,” J. Biomech. 3: 431–451, 1970.
Norman, R. W., M. Sharratt, J. Pezzack, and E. Noble. “A Re-examination of Mechanical Efficiency of Hor-
izontal Treadmill Running,” in Biomechanics, Vol. V-B, P. V. Komi, Ed. (University Park Press, Balti-
more, MD, 1976), pp. 87–93.
Quanbury, A. O., D. A. Winter, and G. D. Reimer. “Instantaneous Power and Power Flow in Body Segments
during Walking,” J. Hum. Mov. Stud. 1: 59–67, 1975.
Ralston, H. J. and L. Lukin. “Energy Levels of Human Body Segments during Level Walking,” Ergonomics
12: 39–46, 1969.
Robertson, D. G. E. and D. A. Winter. “Mechanical Energy Generation, Absorption, and Transfer amongst
Segments During Walking,” J. Biomech. 13: 845–854, 1980.
Rudolph, K. S., M. J. Axe, and L. Snyder-Mackler. “Dynamic Stability After ACL Injury: Who Can Hop?”
Knee Surg. Sports Traumatol. Arthrosc. 8(5): 262–269, 2000.
ISTUDY
Saunders, J. B. D. M., V. T. Inman, and H. D. Eberhart. “The Major Determinants in Normal and Pathological
Gait,” J. Bone Joint Surg. Am. 35A: 543–558, 1953.
Whipp, B. J. and K. Wasserman. “Efficiency of Muscular Work,” J. Appl. Physiol. 26: 644–648, 1969.
Williams, K. R. and P. R. Cavanagh. “A Model for the Calculation of Mechanical Power during Distance
Running,” J. Biomech. 16: 115–128, 1983.
Winter, D. A., “Biomechanical Model Relating EMG to Changing Isometric Tension,”. in Dig. 11th Int. Conf.
Med. Biol. Eng., pp. 362–363, 1976.
Winter, D. A. “Energy Assessments in Pathological Gait,” Physiother. Canada 30: 183–191, 1978.
Winter, D. A. “A New Definition of Mechanical Work Done in Human Movement,” J. Appl. Physiol.
46: 79–83, 1979.
Winter, D. A. “Moments of Force and Mechanical Power in Slow Jogging,” J. Biomech. 16: 91–97, 1983.
Winter, D. A., A. O. Quanbury, and G. D. Reimer. “Analysis of Instantaneous Energy of Normal Gait,”
J. Biomech. 9: 253–257, 1976.
Winter, D. A. and D. G. Robertson. “Joint Torque and Energy Patterns in Normal Gait,” Biol. Cybern.
29: 137–142, 1978.
Yack, H. J. “The Mechanics of Two-Joint Muscles During Gait,” PhD Thesis, University of Waterloo,
Waterloo, Ont., Canada, 1986.
ISTUDY
7
UNDERSTANDING 3D KINEMATIC
AND KINETIC VARIABLES
THOMAS HULCHER
Department of Physical Therapy, Jefferson College of Rehabilitation Science, Thomas Jefferson University,
7.0 INTRODUCTION
Over the past 10 years, there have been major commercial advances in three-dimensional (3D)
hardware and software, with the most recent being in markerless technology. Chapter 3 included
descriptions of some of the 3D imaging and software systems that have been introduced. The
majority of the systems are near infrared, with multiple-camera arrangements requiring passive
reflective markers, while other systems use video-based cameras and markerless technology.
Regardless of the system used, the output of the data collection stage is a file of x, y, z coordinates
of each of the markers or each segment in markerless technology at each sampled point in time.
These coordinates are in the global reference system (GRS) that is fixed in the laboratory or data
collection space. The purpose of this chapter is to go through the steps where these coordinate data
are transformed into the anatomical axes of the body segments so that a kinetic analysis can be done
in a similar manner, as has been detailed for two-dimensional (2D) analyses in previous chapters.
7.1 AXES SYSTEMS
There are several axes reference systems that must be introduced in addition to the GRS. The
markers that are placed on each segment provide a marker axis system that is a local reference
system (LRS) for each individual segment. A second LRS is the axis system that defines the prin-
cipal axis of each segment. Because skeletal landmarks are used to define these axes, this system
is referred to as the anatomical axis system.
7.1.1 Global Reference System

For purposes of convenience, we will be consistent in our axis directions for the GRS: X is
the forward/backward direction, Z is the vertical (gravitational) axis, and Y is the left/right

Companion website:
ISTUDY
146 UNDERSTANDING 3D KINEMATIC AND KINETIC VARIABLES
(medial/lateral) axis. Thus, the XY plane is the horizontal plane and, by definition, is orthogonal to
the vertical axis. The directions of these GRS axes are the same as those of the axes in the force
plate. To ensure that this is so, a spatial calibration system (a rigid L-frame of precisely known
dimensions or a rigid 3D mechanical axis) is instrumented with markers and is placed on one of
the force plates and aligned along the X and Y axes of the force platform. The position of each of
the markers relative to the origin of the force plate is known and fed into the computer. If more
than one force platform is used, the origin of each additional platform is recorded by an X and Y
offset from the primary platform. An additional offset in the Z direction would be necessary if the
additional platform were at a different height from the first (as would be necessary for a biome-
chanical analysis of stairway or ramp walking). In many research situations, the cameras are rear-
ranged to best capture the movement of a particular study and, therefore, require a new calibration
of the GRS each time they are moved. Once the calibration is complete, the cameras cannot be
moved, and care must be taken to ensure they are not accidentally displaced. Even in laboratories
where the equipment is not moved, it is important to periodically calibrate the cameras to prevent
drift and error propagation.
7.1.2 Local Reference Systems and Rotation of Axes

Within each segment, the anatomical axis system is set with its origin at the center of mass (COM)
of the segment, and its principal z axis is usually along the long axis of the segment or, in the case
of segments like the pelvis, along a line defined by skeletal landmarks such as posterior superior
iliac spine and anterior superior iliac spine (PSIS and ASIS). The other local axis system is con-
structed on the segment by using a set of surface markers. A total of two transformations are
necessary to get from the GRS to the marker axis system and from the marker to the anatomical
axis system. Figure 7.1 shows how one of those rotations is done. The axis system x, y, z needs to
be rotated into the system denoted by x , y , z . Many sequences of rotation are possible; here,
we use the common Cardan sequence x–y–z. This means that we rotate about the x axis first, about
the new y axis second, and about the new z axis last. The first rotation is θ1 about the x axis to get
x , y , z . Because we have rotated about the x axis, x will not be changed and x = x, while the y axis
changes to y and the z axis to z . The second rotation is θ2 about the new y axis to get x , y , z .
Because this rotation has been about the y axis, y = y . The final rotation is θ3 about the new z
zʹʺ zʹ
xʹʺ
zʺ θ1
θ2 θ3 xʺ
θ2
xʹ
x
θ1
θ3
y yʹ
yʹʺ
yʺ
Figure 7.1 Cardan sequence of three rotations about the x, y, z axes. The first rotation, θ1, about the x axis to get x ,
y , z ; the second rotation, θ2, about the new y axis to get x , y , z , and the final rotation, θ3, is about the new z axis
to get the desired x , y , z .
ISTUDY
7.1 AXES SYSTEMS 147
axis to get the desired x , y , z . Assuming that we have a point with coordinates x0, y0, z0 in the
original x, y, z axis system, that same point will have coordinates x1, y1, z1 in the x , y , z axis
system. Based on the rotation θ1:
x1 = x0
y1 = y0 cos θ1 + z0 sin θ1
z1 = − y0 sin θ1 + z0 cos θ1
Using the shorthand notation c1 = cosθ1 and s1 = sinθ1, in matrix notation, this may now be
written as:
x1 1 0 0 x0 x0
y1 = 0 c1 s1 y0 = Φ1 y0 (7.1)
z1 0 − s1 c1 z0 z0
After the second rotation θ2 about y , this point will have coordinates x2, y2, z2 in the x , y , z
axis system.
x2 c2 0 − s2 x1 x1
y2 = 0 1 0 y1 = Φ2 y1 (7.2)
z2 s2 0 c2 z1 z1
Finally, the third rotation θ3 about z yields the coordinates x3, y3, z3 in the x , y , z axis
system.
x3 c3 s3 0 x2 x2
y3 = − s3 c3 0 y2 = Φ3 y2 (7.3)
z3 0 0 1 z2 z2
Combining Equations (7.1)–(7.3), we get:
x3 x0
y3 = Φ3 Φ2 Φ1 y0 (7.4)
z3 z0
Note that the matrix multiplication as shown in Equation (7.4) is not commutative, which
means that the order of the transformations must be such that [Φ1] is done first, [Φ2] second,
and [Φ3] last. In other words, [Φ1][Φ2] [Φ2][Φ1] An expansion of Equation (7.4) yields:
x3 c2 c3 s3 c 1 + s 1 s2 c 3 s1 s3 − c1 s2 c3 x0
y3 = − c2 s3 c1 c3 − s1 s2 s3 s1 c3 + c1 s2 s3 y0 (7.5)
z3 s2 − s1 c2 c1 c2 z0
7.1.3 Other Possible Rotation Sequences

In theory, there are 12 possible correct rotation sequences; all were introduced by the Swiss math-
ematician, Leonhard Euler (1707–1783). The list that
ISTUDY
sequences. The example explained previously is generally referred to as the Cardan system,
which is commonly used in biomechanics, while the z–x–z rotation sequence, generally referred
to as the Euler system, is commonly used in mechanical engineering.
x−y −x x − y − z Cardan x−z −x x−z −y

y−x −y y−x −z y−z −x y−z −y
z−x −y z − x − z Euler z−y −x z−y −z
7.1.4 Dot and Cross Products

In three-dimensional motion analysis, we are dealing almost exclusively with vectors. When vec-
tors are multiplied, we must compute the mathematical function called the dot or cross product.
The dot product is also called the scalar product because the result is a scalar while the cross
product is also called the vector product because the result is a vector. Dot product was first intro-
duced in Chapter 6 in the calculation of the mechanical power associated with a force and velocity
vector. Only the component of the force, F, and the velocity, V, that are parallel result in the
power, P = F V = F V cos θ where θ is the angle between F and V in the FV plane. In 3D,
the power, P = F x V x + F y V y + F z V z .
Cross products are used to find the product of two vectors in one plane where the product is a
vector normal to that plane. Suppose we have vectors A = (Ax, Ay, Az) and B = (Bx, By, Bz) and
the cross product is a vector C defined as C = (A × B). C is perpendicular to A and B in the
direction defined by the right-hand rule and has a magnitude C = A B sin θ where θ is the angle
between A and B. The vector C is calculated by the expansion of the determinant where i, j, and k
are unit vectors in the x, y, z axes:
i j k
C = A × B = Ax Ay Az
Bx By Bz
C = Ay Bz − Az By i − Ax Bz − Az Bx j + Ax By − Ay Bx k
(7.6)
= iC x + jC y + kC z
7.2 MARKER AND ANATOMICAL AXES SYSTEMS
The following description outlines the steps that are necessary to transform the x, y, z marker
coordinates from the GRS to the anatomical axes of the segments of the person whose movement
is being analyzed. Figure 7.2 presents the axis systems involved for a given segment whose COM
is at c and whose axes x–y–z are as shown. The GRS has axes X–Y–Z, and they are fixed for any
given camera arrangement. The second axis system, xm–ym–zm, is the marker axis system for each
segment, and this can vary from laboratory to laboratory. Even within a given laboratory, each
experiment could have a different arrangement of markers. For a correct 3D analysis, there must
be at least three independent markers per body segment. The markers on each segment must not
be collinear, which means they must not be in a straight line. They must form a plane in 3D space;
as shown in Figure 7.2, the three tracking markers mT1, mT2, and mT3 define the tracking marker
plane. This plane is assumed to contain the xm and zm axes such that all three markers are in the
+ve xm and the +ve zm quadrant. One point on this marker plane is arbitrarily chosen as the origin
of the marker axes system; here mT1 is chosen and is labeled
ISTUDY
7.2 MARKER AND ANATOMICAL AXES SYSTEMS 149
m z
mC2
Z zm
mT3
Anatomical axes
c x xm
c
Marker axes
Rc y m mT1
mC1
Rm mT2
X
Global axes
ym
Y
Global axes
Markers
G to M matrix
Marker axes
M to A matrix G to A matrix
Calibration Constant
points
Anatomical axes θ1, θ2, θ3
Figure 7.2 An anatomical segment showing the GRS, the marker axes, and anatomical axes. Three tracking
markers, mT1, mT2, and mT3, in conjunction with two calibration markers, mC1 and mC2, define the constant [M
to A] matrix. The three tracking markers in the GRS define the [G to M] matrix. The product of the variable
[G to M] and the fixed [M to A] matrices gives the variable [G to A] matrix from which a new θ1, θ2, and θ3
are defined for each frame.
the +ve zm axis; ym is normal to the tracking plane, and xm is orthogonal to the plane defined by
ym–zm to form a right-hand system.
The purpose of the anatomical calibration process is to find the relation between the marker
axes, xm–ym–zm, and the anatomical axes, x–y–z. This process requires the subject to assume a
well-defined position; usually, the anatomical position is used or a “T” position. At this time,
extra calibration markers may be placed temporarily on the segment to define well-known ana-
tomical points from which the segment’s anatomical axes can be defined. For example, for the leg
segment, the three tracking markers could be placed on the head of the fibula (mT3), on the lateral
malleolus (mT2), and at the midpoint on the anterior surface of the tibia (mT1). During calibration,
temporary markers, mC1 and mC2, could be placed on the medial malleolus and the medial epi-
condyle of the tibia, respectively. With the subject standing still for about a second, the coordi-
nates of the three tracking and the two calibration markers are recorded and averaged over the
calibration time. The long axis of the leg segment (z axis) would then be defined as the line join-
ing the midpoint between the lateral and medial malleoli (mT2 and mC1) to the midpoint between
the head of the fibula and the medial epicondyle of the tibia (mT3 and mC2). These midpoints are
the ankle and knee joint centers, respectively. The leg z axis
ISTUDY
plane normal to which is the leg x axis. The direction of the leg y axis would be defined as a line
normal to the leg x–z plane such that the leg x–y–z is a dextral system. The anatomical axes of the
leg are now defined relative to the three tracking markers. The location of the COM of the leg
would be a known distance along the z axis of the leg from the ankle joint; thus, the vector, c, from
m, the origin of the tracking marker axis system, is also known. The two calibration markers are
now removed and are no longer needed because the orientation of the axis system of the three
tracking markers is now known and assumed to be fixed relative to the newly defined anatom-
ical axes.
In clinical gait laboratories, it is impossible for many patients, such as patients with cerebral
palsy or previous stroke, to assume the anatomical position for even a short period of time. Thus,
the clinical gait teams have developed a consistent marker arrangement combined with a number
of specific anthropometric measures. These include such measures as ankle and knee diameters
which, when combined with generic X-ray anthropometric measures, allow the team to input an
algorithm so that offset displacements from the tracking markers to the joint centers are known.
Patients are then asked to assume a static standing position with a number of temporary calibra-
tion markers similar to what has been described previously. In effect, the major single difference
in the clinical laboratory is that calibration is performed on the patient in a comfortable standing
position rather than the anatomical position. For the complete detailed steps in arriving at the joint
kinetics in an operational clinical laboratory, the reader is referred to Davis et al. (1991) and
Õunpuu et al. (1996).
In Figure 7.2, we see two matrix rotations. [G to M] is a 3 × 3 rotation matrix that rotates from
the GRS to the tracking marker axes, xm–ym–zm. This is a time-varying matrix because the track-
ing marker axes will be continuously changing relative to the GRS. [M to A] is a 3 × 3 matrix that
rotates from the tracking marker axes to the anatomical axes. This matrix is assumed to be con-
stant and results from the calibration protocol. The combination of these two rotation matrices
gives us the [G to A] rotation matrix, which, when solved for a selected angle sequence, yields
the three time-varying rotation angles, θ1, θ2, θ3. With this final matrix, we can get the orientation
of the anatomical axes directly from the tracking marker coordinates that are collected in
the GRS.
However, from Figure 7.2, we are not yet finished. We also have to find a translational trans-
formation to track the 3D coordinates of the segment COM, c, over time. The location of c is
defined by the vector, Rc, which is a vector sum of Rm + c. The vector Rm is the GRS coordinates
of the tracking marker, mT1, while c is a constant vector joining m to c, as defined earlier.
7.2.1 Markerset Design

The same type of markers is typically used for both tracking and calibration. These markers are
usually spherical, though cubic markers have been used in the past, and can vary in size from
1 mm to 5 cm in diameter. The appropriate marker size is determined by: (1) proximity of the
subject to the camera and (2) the resolution of the camera. In the past, camera resolution was
much lower than modern cameras requiring the large 5 cm markers. Current cameras have higher
resolution and therefore much smaller markers can be used. For most applications, a marker that
is 10 mm in diameter is appropriate. Markers under 5 mm are commonly used when the camera is
in close proximity to the participant and a large number of markers are required, such as in facial
expression tracking.
When selecting a marker size, marker density needs to be considered. Each marker needs to be
seen as a distinctive point and caution should be taken when placing a marker close to another.
Depending on the camera’s view, the two markers may be detected as a single marker. Motion
capture cameras work by detecting reflected near-infrared light. When a reflection is seen, it finds
the centroid of that reflection and uses that location as the center of the marker, as shown in
Figure 7.3a. If multiple markers are overlapping in the camera view, the computer will see
the markers as one large reflection and find the centroid of
ISTUDY
(a)
(b)
Figure 7.3 A pictorial representation of how a camera sees and interprets a (a) single and (b) overlapping marker.
The black square in each is the interpreted centroid of the viewed reflection.
marker, as shown in Figure 7.3b, resulting in neither marker having its location correct and a
marker missing from the markerset. The closer a marker is to the camera, the more pixels are
dedicated to its reflection and, conversely, the farther away it is, the fewer pixels are dedicated
to it. Therefore, choosing a marker that will be represented by an adequate number of pixels, but
not overlap with other markers, is important to the markerset selection process.
In addition to size, there are other considerations when selecting an appropriate marker for
motion capture. One of the biggest considerations is how the marker will be attached to the sub-
ject. Commercially sold markers generally come with a flat base attached to the sphere and many
researchers use hoop-and-loop fasteners (like Velcro) to easily attach and remove the markers
from subjects. In scenarios where the markers will always be in the same location, such as on
a piece of sporting equipment, the markers can also be glued to the location. The markers must
be reflective to the radiation emitted by the cameras (typically near infrared) and capable of
withstanding wear and tear. When testing high-speed motions that are often observed in athletes,
markers can collide with objects or become detached and fly off the participant. In these
situations, double-sided tape or strong adhesive tape is recommended to surround the marker base
to prevent detachment even when the athlete begins sweating. The marker coatings must be
periodically inspected for damage to the reflective surface, such as nicks and tears. When this
damage is found, the markers should be repaired with new reflective coating or replaced because
breakdown of the reflective surface inhibits the ability of the marker to be properly tracked.
Over time some standardized markersets have been created and commonly used or modified to
collect biomechanics data. The earliest and most well-known markersets are the Helen-Heyes and
Cleveland Clinic markersets, which are full-body tracking markersets. The Helen-Heyes marker-
set uses a minimal number of markers to define and track the segments. However, because of the
minimal nature of the markerset, it has more constraints on its degrees of freedom available for
calculation and a higher error rate in joint angle calculations (Collins et al., 2009). Many research
articles have used this markerset or a modified version in defining the body segments. The Cleve-
land Clinic markerset uses clusters on the lower body segments for tracking in order to better
display data in the transverse plane. In response to the error issues with the Helen-Heyes and
Cleveland Clinic markersets, a six degrees of freedom (6DOF) markerset was developed, com-
bining the strong elements of the previous two markersets,
ISTUDY
clusters on the thigh and shank segments in order to better track motion in all planes (Collins
et al., 2009). In many labs, researchers use these markersets as a starting point to create a mar-
kerset that best captures the data important to the research question under investigation and one of
the strengths of markersets is the flexibility they provide in researching movement.
When placing the markers or collecting data, issues can arise that researchers need to be cog-
nizant of and know how to fix. One of the most common problems is the medial markers, espe-
cially on the legs, colliding with each other and falling off. This can be fixed by ensuring there are
enough tracking markers (at least 3) on each segment so that the medial markers are only needed
in calibration and can be removed for dynamic collection. When using marker clusters on a solid
backing, the clusters are frequently affixed to cloth wrapped around the segment. Attention needs
to be paid to these cloth wraps to make sure they are not sliding down the limb as time progresses,
which will create significant errors as the markers shift from the calibration reference position.
7.2.2 Event Detection Methods for Gait

One of the most common activities researched in biomechanics is gait. Gait data is often broken
down into the stance phase and swing phase and the events that define the start and end of these
phases are (usually) heel contact (HC) and toe-off (TO). These events are used as temporal mar-
kers to align data and compare biomechanical measures across trials and participants, such as
graphing the knee angle during the stance phase (from HC to TO). Most often, these gait events
are determined using force plates in conjunction with motion capture data. HC and TO can be
defined as the time point in which the force measurement rises above and falls below a threshold
value, respectively. This threshold is often set at 20 N to avoid events being placed due to noise or
the foot brushing the force plate in a low clearance swing. This technique works well in healthy
participants with typical HC and TO kinematics. However, in participants, such as those with
cerebral palsy, with atypical HC and TO kinematics, this technique often fails.
Therefore, computing HC and TO in participants with atypical HC and TO kinematics or when
kinetic data is not available can be challenging. One method that has been shown to work con-
sistently is to find when the maximal anterior/posterior displacement between the heel marker and
the sacrum marker occurs, which will determine HC. TO will be determined when maximal
anterior/posterior displacement between the toe marker and the sacrum marker occur (Zeni
et al., 2008). Another method, if the subject is on a treadmill, is to use the anterior/posterior veloc-
ity of the toe or heel marker. The instant the marker changes from positive to negative is when HC
occurs and the instant the marker changes from negative to positive is when TO occurs (Zeni
et al., 2008). While these algorithms work well for gait at normal walking speeds, they may
not be accurate during running analyses.
When collecting data that will be using these methods to determine gait events, certain atten-
tion needs to be paid in collection. When using kinetic data, careful observation of foot placement
is required. Because a force threshold is used as the trigger for an event to be created, the
researcher must make sure that the foot was fully on one force plate exclusively and no part
of the foot was in contact with any other surface at any point during stance. Additionally, no other
forces can be applied to the force plate, such as objects, other feet, canes, etc. Attention needs to
be paid during collections to the force plates to ensure they remain properly tared so that electrical
noise or drift does not create gait events that are not real. The kinematic data gait event methods
are also especially vulnerable to nonstandard gait patterns and may not be valid if the subject has
gait deviations or other disabilities that alter standard biomechanics.
Once data has been collected for a participant, it is often compared to others to determine differ-
ences in biomechanical variables. First, the kinematic and kinetic data should be time normalized to
100% of the stance and/or swing phase so that the variables are temporally aligned. In order to
control for varying body size, normalization techniques should be applied to kinetic data as well.
Each kinetic variable may have different considerations; therefore, determination of appropriate
normalization techniques should be done on a
ISTUDY
normalize kinetic data are (1) divide by the mass of the participant, (2) by the height of the partic-
ipant, or (3) by the velocity of the participant. Examining standard practice among fields of study is
a good way to determine the best to way to normalize the kinetic data for comparison.
7.2.3 Event Detection Methods for Other Activities

Although gait is the most common motion studied in biomechanics, there are many other motions
and activities that are of interest, including jumping/landing and squatting. Jumping and landing
events often divide the motion into jump initiation, liftoff, touchdown, and completion of land-
ing. Jump initiation and landing completion can be detected when the participant deviates from a
standing posture and returns to a standing posture, respectively. This is often done by setting a
time event when a marker, usually on the pelvis, first appreciably moves in the vertical direction
and returns to the full standing height after the touchdown event. The liftoff and touchdown time
points are easiest to determine with the use of force thresholding using the vertical force com-
ponent of the force plate data.
Squatting is another activity that is frequently examined in research. It is often divided into the
lowering and standing phase. These events are set similarly to the jump initiation and landing
completion events from jumping, where the maximum and minimum vertical displacement of
the pelvic marker determines the lowering and standing phases of the squat.
While the above provides basic examples of event detection techniques, it is by no means
exhaustive. The events are often movement specific and need to be determined based on the spe-
cific analysis considering the markerset used, camera set-up, and equipment.
7.2.4 Considerations for Applications with Implements

In biomechanics research, especially in the context of sports or for individuals who need assis-
tance walking, extra implements may be included in the data collection. These implements could
include balls, golf clubs, lacrosse sticks, canes, or walkers. When these implements are included,
it is generally either because, it is a requirement of the sport or activity, the movement could not
be performed without them or the subject is too unstable. For example, to study the biomechanics
of a golf swing it is important that the athlete is swinging the club since the inertial properties of
the club will influence both the kinematics and kinetics of the swing. Therefore, it would be
important to track the golf club during the swing, to calculate the inertial influence on the body.
If it is determined that the implement is important, it should have tracking markers placed on it in
such a way that the movement in all planes is detectable, very similar to the principals involved in
tracking limb segments. While tracking a relevant object may be straightforward, getting useful
data from it may not. Each body segment’s inertial properties have been determined from cadav-
eric studies. (How these inertial properties will be used in calculations is discussed later in this
chapter.) This allows for quick computation because the inertial properties can be inputted in
motion tracking analysis software. Objects like a baseball or basketball are not very challenging
to analyze because they are uniform spheres with uniformly distributed mass and have easily cal-
culated inertial properties. Unfortunately, the inertial properties of something like a golf club
would need to be determined before calculations can be performed on it and is much more com-
plex. Golf clubs, and similar objects, with very irregular shapes have a large portion of mass
located at the extreme end, making the calculation of its inertial properties challenging. Before
conducting research using such objects, its properties will need to be determined in order for it to
be integrated into the model.
One additional thing to consider when using implements is their contact with the floor, espe-
cially if using force plates. The implements can move the distribution of weight of the subject
onto the object and affect the location of the center of pressure of the force on the force plate.
This is prevalent in assistive devices like canes and
ISTUDY
how force distribution and biomechanics changes when using a device, it is recommended to use
alternative methods of support, like a bodyweight harness or gait belt.
7.2.5 Example of a Kinematic Data Set

7.2.5.1 Calibration – Calculation of [Marker to Anatomical] Matrix. Let us now look at an
example of numerical data to see how the various transformations are calculated. The leg segment
in Figure 7.2 will be used as an example. Recall that there are three tracking markers on this seg-
ment plus two calibration markers whose coordinates were digitized during the calibration period
when the subject stood steady in the anatomical position. In Figure 7.2, the subject would be
facing the +ve X direction, and the left leg was being analyzed. Table 7.1 gives the X, Y, and
Z coordinates in the GRS, averaged over one second.
In this calibration position, the ankle = (mT2 + mC1)/2, Xa = 2.815, Ya = 22.685, Za = 10.160.
The coordinates of the knee = (mT3 + mC2)/2, Xk = 6.670, Yk = 20.965, Zk = 41.890. The COM
of the leg = 0 567 × knee + 0 433 × ankle, Xc = 5.001, Yc = 21.710, Zc = 28.151.
Now we have to locate the anatomical x, y, and z-axes. Let the line joining the ankle to the knee
be the z axis, and the line joining the lateral malleolus to the medial malleolus be an interim y axis
(because it is not exactly normal to the y axis but will be corrected later). These two axes now
form a plane, and the x axis, by definition, is normal to the yz plane and, therefore, is the “cross
product” of y and z, or xan = (yan × zan). Using the subscript “an” to indicate anatomical axes:
zan = knee − ankle xz = 3 855, yz = − 1 720, zz = 31 730
yan = mC1 − mT2 xy = − 0 210, yy = 7 670, zy = 0 120
xan = yan × zan xx = 246 653, yx = 7 126, zx = − 29 581
One final correction must be done to our anatomical model. yan is the line joining the medial to
lateral malleoli and is approximately 90 from the long axis zan. To ensure that all three anatom-
ical axes are at right angles to each other, yan must be corrected: yan = (xan × zan): xy = −175.229,
yy = −7940.334, zy = −451.714. Note that none of these vectors is unity length, and they are nor-
mally reported as a unit vector. For example, the length of xan is 245.423 cm; thus, dividing each
coordinate by the length of xan yields a unity vector for xan: xx = 0.9925, yx = −0.1190, zx = 0.0290.
Similarly, as a unit vector, the corrected yan is: xy = −0.0274, yy = 0.9995, zy = 0.0156. Similarly,
for zan as a unit vector, xz = 0.1204, yz = −0.0537, zz = 0.9913. We can now construct the leg
anatomical-to-global matrix [LA to G] and it is:
0 9925 − 0 0274 0 1204

− 0 1190 0 9995 − 0 0537
0 0290 0 0156 0 9913
TABLE 7.1 Marker Coordinates During Standing Calibration

Marker Location X (cm) Y (cm) Z (cm)
mT1 Midleg 9.39 21.90 30.02
mT2 Lateral malleolus 2.92 18.85 10.10
mT3 Fibular head 5.05 15.41 41.90
mC1 Medial malleolus 2.71 26.52 10.22
mC2 Medial condyle 8.29 26.52 41.88
ISTUDY
Note that the diagonal values are almost = 1, indicating that the subject stood in the calibration
position with his three anatomical axes almost perfectly lined up with the global axes. A more
useful transformation matrix is the leg global-to-anatomical [LG to A], which is the transpose of
[LA to G]:
0 9925 − 0 1190 0 0290

− 0 0274 0 9995 0 0156
0 1204 − 0 0537 0 9913
These anatomical axes have their origins at the ankle joint. However, from our inverse
dynamics, it is more convenient that the origin of the anatomical axes (see Figure 7.2) be
located at the COM of the leg segment with coordinates 0, 0, 0. We must now establish the
local coordinates of the ankle and knee joints and the three tracking markers relative to this
new origin at the COM.
From the COM to the ankle vector = (global ankle − global COM),
xal = X a − X C = 2 815 − 5 001 = − 2 186,

yal = Y a − Y C = 22 685 − 21 710 = 0 975,
zal = Z a − Z C = 10 160 − 28 151 = − 17 991
The anatomical ankle vector is the product of [LG to A][ankle vector]:
0 9925 − 0 1190 0 0290 − 2 186 − 2 81

− 0 0274 0 9995 0 0156 0 975 = 0 75
0 1204 − 0 0537 0 9913 − 17 991 − 18 15
This anatomical ankle vector lies along the line joining the ankle to knee and the ankle is 18.15
cm distal of the COM. If we were to repeat this procedure for the anatomical knee vector and for
the anatomical vectors of the three tracking markers mT1, mT2, and mT3, we would calculate the
following:
Anatomical knee Anatomical mT1 Anatomical mT2 Anatomical mT3

vector vector vector vector
2 14 4 39 − 2 25 1 20
− 0 57 0 10 − 3 08 − 6 08
19 86 2 37 − 17 99 13 97
We are now ready to calculate the constant marker-to-anatomical matrix ([M to A] in

Figure 7.2). The three tracking markers form a plane in the GRS, and we can now define our
maker axes in that plane. mT1 is chosen as the origin of the marker plane, and the line joining
mT1 to mT3 is chosen to be the z axis, labeled zm. The line joining mT2 to mT1 is a vector labeled
A (an interim vector to allow us to calculate ym and xm). ym is normal to the plane defined by zm
and A and xm, is normal to the plane defined by ym and zm.
zm = anatomical mT3 − anatomical mT2 3 45, − 3 00, 31 96

A vector = anatomical mT1 − anatomical mT2 6 64, 3 18, 20 36
ym = zm × A vector − 162 71, 141 97, 30 89
ISTUDY
x m = y m × zm 4630 11, 5306 88, − 1 67
The normalized axis for this leg anatomical-to-marker matrix [LA to M] is:
0 6574 0 7535 − 0 0002

− 0 7459 0 6508 0 1416
0 1069 − 0 0929 0 9899
The fixed leg marker-to-anatomical matrix [M to A] is the transpose of [LA to M]:
0 6574 − 0 7459 0 1069

0 7535 0 6508 − 0 0929
− 0 0002 0 1416 0 9899
7.2.5.2 Tracking Markers – Calculation of [Global to Marker] Matrix. We are now ready to
calculate the [G to M] matrix in Figure 7.2. Table 7.2 lists representative GRS coordinates for the
leg segment for three successive frames of walking taken during the swing phase. The procedure
to calculate this [G to M] matrix is exactly the same as the latter part of the calculation of the [M to
A] matrix. Consider the coordinates for frame 6:
zm = mT3 − mT2 24 34, − 3 64, 19 99
A vector = mT1 − mT2 18 86, 4 09, 9 15
ym = zm × A vector − 115 07, 154 30, 168 20
x m = y m × zm 3696 72, 6394 16, − 3336 83
The normalized axis for this leg [G to M] matrix for frame 6 is:
0 4561 − 0 4502 0 7676

0 7889 0 6036 − 0 1148
− 0 4117 0 6580 0 6304
TABLE 7.2 Tracking Markers During Walking

mT1 mT2 mT3
Frame X Y Z X Y Z X Y Z
5 20.65 35.95 33.87 1.30 32.14 25.74 26.52 28.10 44.43
6 25.46 35.95 34.47 6.60 31.86 25.32 30.94 28.22 45.31
7 30.18 35.94 34.97 11.98 31.60 24.64 35.08 28.36 46.10
ISTUDY
7.2.5.3 Calculation of [Global to Anatomical] Matrix. From Figure 7.2, the final step is
to calculate the [G to A] matrix that is the product of the fixed [M to A] matrix and the variable
[G to M] matrix; for frame 6 this product is:
0 6574 − 0 7459 0 1069 0 4561 − 0 4502 0 7676
0 7535 0 6508 − 0 0929 0 7889 0 6036 − 0 1148
− 0 0002 0 1416 0 9899 − 0 4117 0 6580 0 6304

− 0 3326 − 0 6758 0 6576
= 0 8953 − 0 0075 0 4451
− 0 2959 0 7369 0 6076
From Equation (7.5), this [G to A] matrix is equal to:
c2 c3 s3 c1 + s1 s2 c3 s1 s3 − c1 s2 c3
− c2 s3 c1 c3 − s1 s2 s3 s1 c 3 + c 1 s2 s3
s2 − s1 c2 c1 c2
We now solve this matrix to get θ1, θ2, and θ3. Equating the three terms in the bottom row:
s2 = −0.2959, −s1c2 = 0.7369, c1c2 = 0.6304. Therefore θ2 = −17.21 or 162.79 ; assuming
θ2 = −17.21 , c2 = 0.95522 or −0.9522. Because c1c2 = 0.6304, c1 = 0.6304/
0.95522 = 0.6600, and θ1 = 48.7 or −48.7 . Since −s1c2 = 0.7369 and c2 = 0.95522,
s1 = 0.7369/−0.95522 = 0.7714 or −0.7714. We now validate that θ1 = −48.7 because
−s1c2 ≈ 0.7714. We now use the first two terms in the first column to calculate and validate
θ3: c2c3 = −0.3326, −c2s3 = 0.7535. From this, c3 = −0.3326/0.95522 = 0.3482, θ3 = 69.62
or −69.62 , and s3 = 0.7535/−0.95522 = 0.7888 or −0.7888. The only valid solution is
θ3 = −69.62 because −c2s3 ≈ 0.7500. To summarize the results of these three rotations (see
Figure 7.1), to bring the global axes in line with the anatomical axes requires an initial rotation
about the global X axis of −48.7 . This will create new Y and Z axes and will be followed by a
rotation of −17.21 about the Y axis. This rotation creates new X and Z axes. The final rotation
is the largest (because we are analyzing the leg segment during swing), and it is −69.62 , which
creates the final X , Y , and Z axes. These final axes are the anatomical x–y–z axes shown in
Figure 7.2.
Finally, to get the COM of the segment, we must calculate c in GRS coordinates. We have c in
the leg anatomical reference, and it is composed of [A to G] = [G to A]T and [Anatomical mT2
vector] = [−2.25, −3.08, −17.99]. In the GRS,
c = A to G − 2 25, − 3 08, − 17 99
− 0 3326 − 0 6758 0 6576 − 2 25 3 314
= 0 8953 − 0 0075 0 4451 − 3 08 = − 11 713
− 0 2959 0 7369 0 6076 − 17 99 − 13 781
ISTUDY
From Figure 7.2, the global vector (remember that Rm is equal to the GRS coordinate
of mT1)
Rc = Rm + c = 28 77, 24 24, 20 69
7.2.5.4 ISB Recommendations. When creating segments for biomechanical analysis, it is

important to define the coordinate systems in a standardized format so future studies can compare
data and to reduce mathematical error. The International Society of Biomechanics has published
recommendations on how to define joint segments and the Euler rotation sequence to use in anal-
ysis (Table 7.3).
The ISB also created recommendations for joint segment Euler rotation sequences specifi-
cally for the upper body (Table 7.4). The upper body can often be challenging due to a greater
overall range of motion in multiple planes, which make selection of the rotation sequence less
straightforward than the lower extremity. When certain rotations are used in the upper body the
extremes of motion can lead to mathematical errors such as gimbal lock. Gimbal lock occurs
when two axes of two segments are oriented parallel to each other thereby being unable to
determine the 3D joint position. For the lower body segments, the recommended rotation order
is X–Y–Z.
7.3 DETERMINATION OF SEGMENT ANGULAR VELOCITIES

AND ACCELERATIONS
Recall from Section 7.1.2 and Figure 7.2 that we had to determine three time-varying rotation
angles, θ1, θ2, and θ3, prior to transforming from the GRS to the anatomical axes. The first time
derivative of these transformation angles yields the components of the segment angular
velocities:
ω = dθ1 dt ex + dθ2 dt ey + dθ3 dt ez (7.7a)
where ex , ey and ez denote the unit vectors of the three rotation axes x, y , and z shown in
Figure 7.1. Consider an angular velocity, ω , about axis x; here, ω = dθ1/dt ex and there is
no rotation of θ2 or θ3. This angular velocity can be expressed as:
θ1
ω = 0
0
The second angular velocity ω = dθ2 dt ey , plus the component of ω that is transformed by
[Φ2] in Equation (7.2), can be expressed as:
0 c2 0 − s2 θ1 0 c2 θ 1 c2 θ 1
ω = θ2 + 0 1 0 0 = θ2 + 0 = θ2
0 s2 0 c2 0 0 s2 θ 1 s2 θ1
ISTUDY
TABLE 7.3 ISB Recommendations on Joint Segment Coordinate Systems
Segment Origin x axis y axis z axis
Tibia/ The point halfway The anterior pointing line The line perpendicular to both x and The line connecting the medial and
Fibula between lateral and perpendicular to the frontal plane of z axis lateral malleolus, pointing right
medial malleolus the tibia/fibula
Calcaneus The point halfway The anterior pointing line The line pointing cranially along the The line perpendicular to both x and y
between lateral and perpendicular to the frontal plane of long axis of the tibia/fibula axis
medial malleolus the tibia/fibula
Pelvis The center of rotation The anterior pointing line parallel to the The line perpendicular to both x and The line parallel to the line connecting
of the right hip plane created by the two anterior z axis the right and left anterior superior iliac
superior iliac spine and the midpoint spine, pointing right
of the posterior superior iliac spine
Femur The center of rotation The line perpendicular to both the y and The line pointing cranially through the The line perpendicular to the y axis along
of the right (or left) z axis midpoint of the lateral and medial the plane created by the lateral and
hip femoral epicondyles and the origin medial femoral epicondyles and
origin, pointing to the right
Thorax The point of the The line perpendicular to both the y and The line pointing up connecting the The line perpendicular to the plane
suprasternal notch z axis, pointing forward midpoint of the xiphoid process and formed by the suprasternal notch, the
spinal process of the eighth thoracic spinal process of the seventh cervical
vertebra to the midpoint of the vertebra and the midpoint of the
suprasternal notch and spinal process xiphoid process and the spinal process
of the seventh cervical vertebra of the eighth thoracic vertebra,
pointing right
Clavicle The most ventral point The line perpendicular to both the y The line perpendicular to both x and The line connecting the most ventral
of the axis of the thorax and the z axis of the z axis point on the sternoclavicular joint and
sternoclavicular clavicle the most dorsal point on the
joint acromioclavicular joint, pointing
toward the most dorsal point on the
acromioclavicular joint
Scapula The most laterodorsal The forward pointing line The line perpendicular to both x and The line connecting the trigonum spinae
point of the scapula perpendicular to the plane formed by z axis scapulae and the most laterodorsal
the most caudal point of the scapula, point of the scapula, pointing toward
the most laterodorsal point of the the most laterodorsal point of the
scapula, and the trigonum spinae scapula
scapulae
(continued)
ISTUDY
TABLE 7.3 (Continued)
Segment Origin x axis y axis z axis
Humerus The glenohumeral The forward pointing line The line connecting the glenohumeral The line perpendicular to both x and y
(option rotation center perpendicular to the plane formed by rotation center and the midpoint of axis, pointing right
1) the lateral and medial humeral the lateral and medial humeral
epicondyle and the glenohumeral epicondyle, pointing toward the
rotation center glenohumeral rotation center
Humerus The glenohumeral The line perpendicular to both the y and The line connecting the glenohumeral The line perpendicular to the plane
(option rotation center z axis, pointing forward rotation center and the midpoint of formed by the y axis of the humerus
2) the lateral and medial humeral and the y axis of the forearm, pointing
epicondyle, pointing toward the right
glenohumeral rotation center
Forearm The most caudal- The forward pointing line The proximal pointing line connecting The line perpendicular to both x and y
medial point of the perpendicular to the plane formed by the ulnar styloid and the midpoint of axis, pointing right
ulnar styloid the ulnar styloid, radial styloid and the lateral and medial humeral
midpoint of the lateral, and medial epicondyle
humeral epicondyle
Hand The most caudal- The line perpendicular to both the y and The proximal pointing line along the The line connecting the ulnar and radial
medial point of the z axis long axis of the forearm styloid, pointing right
ulnar styloid
These data are from the ISB recommendations parts 1 and 2 (Wu et al., 2002, 2005).
ISTUDY
7.3 DETERMINATION OF SEGMENT ANGULAR VELOCITIES AND ACCELERATIONS 161
TABLE 7.4 ISB Recommendations on Joint Segment Euler Rotation Sequences and Movement
Directions
Rotation
Segment Sequence x axis y axis z axis
Thorax relative Z–X–Y Lateral flexion rotation Axial rotation Flexion(−)/
to GCS right(+)/left(−) left(+)/right(−) extension(+)
Clavicle Y–X–Z Elevation(−)/ Retraction(−)/ Axial rotation
relative to depression(+) protraction(+) backward(+)/
thorax forward(−)
Scapula Y–X–Z Acromioclavicular Acromioclavicular Acromioclavicular
relative to lateral(−)/ retraction(−)/ anterior(−)/
clavicle medial(+) rotation protraction(+) posterior(+) tilt
Humerus Y(scapula)– Glenohumeral Glenohumeral (Scapula y axis)
relative to X–Y elevation internal(+)/ Glenohumeral plane
scapula external(−) rotation elevation
Scapula Y–X–Z Lateral(−)/ Retraction(−)/ Anterior(−)/
relative to medial(+) protraction(+) posterior(+) tilt
thorax rotation
Humerus Y–X–Y Elevation Internal(+)/ (Thorax y axis) Plane of
relative to external(−) elevation, 0 abduction/
thorax rotation 90 forward flexion
Forearm Z–X–Y Carrying angle Pronation(+)/ Flexion(+)/
relative to supination(−) hyperextension(−)
humerus
These data are from the ISB recommendations parts 1 and 2 (Wu et al., 2002, 2005).
Similarly, the third angular velocity, ω = dθ3 dt ez , plus the contribution from ω that is
transformed by [Φ3] in Equation (7.3), gives us:
0 c3 s3 0 c2 θ 1 0 c 3 c 2 θ 1 + s3 θ 2
ω = 0 + − s3 c3 0 0 = 0 + − s3 c 2 θ 1 + c 3 θ 2
θ3 0 0 1 s2 θ 1 θ3 s2 θ1
c 3 c 2 θ 1 + s3 θ 2
= − s3 c 2 θ 1 + c 3 θ 2
s2 θ 1 + θ 3
Decomposing ω into its three components along the three anatomical axes:
ωx c2 c3 s3 0 θ1
ω= ωy = − c2 s3 c3 0 θ2 (7.7b)
ωz s2 0 1 θ3
We can now calculate the three-segment angular velocities, ωx, ωy, and ωz, that are necessary to
solve the 3D inverse dynamics equations developed in the next section. Recall that the time-varying
θ1, θ2, and θ3 are calculated from Equation (7.5), and the time derivatives of these angles are
individually calculated using the same finite difference technique used in two dimensions. The
three-segment angular accelerations, αx, αy, and αz, can now
difference. We now have all the kinematic variables
ISTUDY
7.4 KINETIC ANALYSIS OF REACTION FORCES AND MOMENTS
Having developed the transformation matrices from global to anatomical and from anatomical to
global, we are now in a position to begin calculating the reaction forces and moments at each of
the joints. Because the ground reaction forces are measured in the GRS and the moments of iner-
tia are known in the anatomical axes, these previously determined transformation matrices are
used extensively in the kinetic calculations. All joint reaction forces are initially calculated in
the GRS, and all joint moments are calculated in the anatomical axes.
7.4.1 Newtonian Three-Dimensional Equations of Motion for a Segment

All reaction forces are calculated in the GRS and, because the gravitational forces and the seg-
ment COM accelerations are readily available in the GRS, it is convenient to calculate all segment
joint reaction forces in the GRS. Students are referred to the 2D link-segment equations and free-
body diagram equations in Section 5.1. Figure 7.4 is now presented to demonstrate the steps
required to calculate kinetics for this segment. The only addition is the third dimension, z.
We are given the three distal reaction forces either as measures from a force plate or from the
Rzp
Mxp
Mzp Rxp
az
Myp z
Ryp
ωx, αx
Ip ωz, αz x
c ax
ay
ωy, αy
Z
Ryd
y
Myd
Id
Rxd
Mzd
Mxd
X
Rzd
Figure 7.4 3D free-body diagram for solution of the inverse dynamics equations. Known are the distal reaction
forces and moments, the COM linear accelerations, and the segment’s angular velocities and accelerations. Using the
kinetic Equations (7.7a), (7.7b), and (7.8a)–(7.8c), we calculate the
ISTUDY
7.4 KINETIC ANALYSIS OF REACTION FORCES AND MOMENTS 163
analysis of the adjacent distal segment. It should be noted that the reaction forces and moments at
the distal end are in the reverse direction from those at the proximal end, the same convention as
was used in Section 5.1.
Step 1: Calculate the reaction forces at the proximal end of the segment in the GRS:
ΣF X = maX or RXP − RXD = maX
ΣF Y = maY or RYP − RYD − mg = maY
ΣF Z = maZ or RZP − RZD = maZ
where aX, aY, aZ are the segment COM accelerations in the X, Y, Z GRS directions and RXP,
RXD, RYP, RYD, RZP, RZD are the proximal and distal reaction forces in the X, Y, and Z axes.
Step 2: Transform both proximal and distal reaction forces into the anatomical axes using the
[G to A] matrix transformation based on θ1, θ2, and θ3 (see Equation (7.5)). We will now
have the proximal and distal reaction forces in the anatomical axes x, y, z: Rxp, Rxd, Ryp, Ryd,
Rzp, Rzd.
Step 3: Transform the distal moments from those previously calculated in the GRS using the
[G to A] matrix to the anatomical axes, as before, based on θ1, θ2, and θ3: Mxd, Myd, Mzd. We
now have all the variables necessary to calculate the proximal moments in the anatomi-
cal axes.
7.4.2 Euler’s Three-Dimensional Equations of Motion for a Segment

The equations of motion for the 3D kinetic analyses are the Euler equations. Considerable sim-
plification can be made in the rotational equations of motion if these equations are written with
respect to the principal (anatomical) axes of the segment with their origin at the COM of the seg-
ment. Thus, the x–y–z axes of the segment in Figure 7.4 satisfy those conditions. The angular
velocity of the segment in its coordinate system is ω. The rotational equations of motion are:
I x αx + I z − I y ωy ωz = ΣM x = Rzd ld + Rzp lp + M xp − M xd (7.8a)
I y αy + I x − I z ωx ωz = ΣM y = M yp − M yd (7.8b)
I z αz + I y − I x ωx ωy = ΣM z = − Rxd ld − Rxp lp + M zp − M zd (7.8c)
where
Ix, Iy, Iz = moments of inertia about x–y–z axes
ωx, ωy, and ωz = components of angular velocity ω about x–y–z axes
αx, αy, αz = components of angular acceleration about x–y–z axes
Mxd, Myd, Mzd = previously transformed distal moments (not shown in Figure 7.4)
about x–y–z axes
Rxd, Rxp, Ryd, Ryp, Rzd, Rzp = previously transformed joint reaction forces about x–y–z axes
lp, ld = distances from COM to proximal and distal joints.
The unknowns in these three equations are the three moments (Mxp, Myp, and Mzp) about the
proximal x, y, z axes. Note that Equations (7.8a)–(7.8c) are in the same form as the 2D
Equation with the additional term (I1 − I2) ω1ω2 to account for the interaction of the angular velo-
cities in the other two axes. Also note that the moments
ISTUDY
do not involve the proximal and distal reaction forces because these forces have zero moment
arms about that axis.
7.4.3 Example of a Kinetic Data Set

The set of kinematic and kinetic data shown in Tables 7.5 and 7.6 is taken
from stance phase of walking, and we will confine our analysis to the leg segment
(see Figure 7.4). The following anthropometric measures apply m = 3.22 kg, Ix = 0.0138 kg
m2, Iy = 0.0024 kg m2, Iz = 0.0138 kg m2, ld = 13.86 cm = 0.1386 m, Ip = 18.15 cm =
0.1815 m ld = 13.86 cm = 0.1386 m, Ip = 18.15 cm = 0.1815 m, frame rate = 60 Hz.
From the kinematic and kinetic measures presented in Tables 7.5 and 7.6, we complete the
analysis of frame 6: aX = 7.029 m/s2, aY = 1.450 m/s2, aZ = − 0.348 m/s2 aX = 7.029 m/s2,
aY = 1.450 m/s2, aZ = − 0.348 m/s2. Refer to the three steps in Section 7.4.1.
Step 1
RXP − RXD = max , RXP = maX + RXD , RXP = 3 22 × 7 029 + − 119 04 = − 96 41 N
RYP − RYD − mg = maY , RYP = maY + RYD + mg, RYP = 3 22 × 1 450 + − 12 03

+ 3 22 × 9 814 = 24 24 N
RZP − RZD = maZ , RZP = maz + RZD , RZP = 3 22 × − 0 348 + − 791 44 = − 792 56 N
Step 2
θ1 = − 0 14781 rad = − 8 46889,
θ2 = − 0 02501 rad = − 1 43297,
θ3 = − 0 46053 rad = − 26 38643
cos θ1 = 0 9891, sin θ1 = − 0 1476

cos θ2 = 0 9997, sin θ2 = − 0 025
cos θ3 = 0 8958, sin θ3 = − 0 4444
TABLE 7.5 Ankle Reaction Forces (N) and Moments (N m) in GRS

Frame RXD RYD RZD MXD MYD MZD
4 −86.55 −13.81 −766.65 10.16 −97.55 6.74
5 −102.71 −12.12 −790.27 11.94 −102.50 4.65
6 −119.04 −12.03 −791.44 14.29 −103.85 2.11
7 −134.33 −10.26 −763.38 16.13 −101.32 −0.49
8 −146.37 −5.26 −704.00 16.64 −94.59 −2.88
TABLE 7.6 Leg Angular Displacements (rad) and Velocities (rad/s)

▪ ▪ ▪
Frame θ1 θ2 θ3 θ1 θ2 θ3
4 −0.14246 −0.02646 −0.37248 −0.1752 0.2459 −2.358
5 −0.14512 −0.02362 −0.41396 −0.1607 0.0434 −2.641
6 −0.14781 −0.02501 −0.46053 −0.0911 −0.1048 −2.909
7 −0.14815 −0.02705 −0.51094 −0.1139 −0.0923 −3.209
8 −0.15161 −0.02809 −0.56749 −0.1109 0.2175 −3.472
ISTUDY
Substituting these values in the [G to A] matrix (Equation (7.5)), we get:
c2 c3 s3 c 1 + s 1 s2 c 3 s1 s3 − c1 s2 c3 0 8955 − 0 4363 0 0875

− c2 s3 c1 c3 − s1 s2 s3 s1 c3 + c1 s2 s3 = 0 4443 0 8877 − 0 1215
s2 − s1 c2 c1 c2 − 0 0253 0 1473 0 9888
We can now transform the reaction forces from the global to the anatomical axes system:
Rxd 0 8955 − 0 4363 0 0875 RXD

Ryd = 0 4443 0 8877 − 0 1215 RYD
Rzd − 0 0253 0 1473 0 9888 RZD
0 8955 − 0 4363 0 0875 − 119 04 − 170 603
= 0 4443 0 8877 − 0 1215 − 12 03 = 32 5915
− 0 0253 0 1473 0 9888 − 791 44 − 781 336
Rxp 0 8955 − 0 4363 0 0875 RXP

Ryp = 0 4443 0 8877 −0 1215 RYP
Rzp − 0 0253 0 1473 0 9888 RZP
0 8955 − 0 4363 0 0875 −96 41 −166 26
= 0 4443 0 8877 −0 1215 24 24 = 74 9789
− 0 0253 0 1473 0 9888 −792 56 −777 674
Step 3: In a similar manner, we transform the ankle moments from the global to the anatomical
axes system:
M xd 0 8955 − 0 4363 0 0875 M XD
M yd = 0 4443 0 8877 − 0 1215 M YD
M zd − 0 0253 0 1473 0 9888 M ZD
0 8955 − 0 4363 0 0875 14 29 58 2911
= 0 4443 0 8877 − 0 1215 − 103 85 = − 86 095
− 0 0253 0 1473 0 9888 2 11 − 13 5723
Using Equation (7.7b), we calculate the angular velocities and accelerations required for the
solution of Euler’s kinetic Equations (7.8a)–(7.8c). As previously calculated for frame 6,
c2 = 0.9997, c3 = 0.8958, s2 = −0.025, s3 = −0.4444.
ωx c2 c3 s3 0 θ1 0 8955 − 0 4444 0 − 0 0911 − 0 0350

ωy = − c 2 s3 c3 0 θ2 = 0 4443 0 8958 0 − 0 1048 = − 0 1344
ωz s2 0 1 θ3 − 0 0253 0 1 − 2 909 − 2 7067
ISTUDY
Similarly, for frame 5,

ωx − 0 1632
ωy = − 0 0325
ωz − 2 6369
and for frame 7,

ωx − 0 0610
ωy = − 0 1333
ωz − 3 2061
We now calculate the angular accelerations of the segment where Δt is the sampling period:
αx fr 6 = ωx fr 7 − ωx fr 5 2Δt = − 0 0610 − − 0 1632 2 × 1 60 = 3 066 rad s2
αy fr 6 = ωy fr 7 − ωy fr 5 2Δt = − 0 1333 − − 0 0325 2 × 1 60 = − 3 024 rad s2
αz fr 6 = ωz fr 7 − ωz fr 5 2Δt = − 3 2061 − − 2 6369 2 × 1 60 = − 17 076 rad s2
We are now ready to solve Euler’s Equations (7.8a)–(7.8c) for the proximal moments:
I x αx + I z − I y ωy ωz = ΣM x = Rzd ld + Rzp lp + M xp − M xd
M xp = I x αx + I z − I y ωy ωz − Rzd ld − Rzp lp + M xd
= 0 0138 × 3 066 + 0 0138 − 0 0024 − 0 1344 × − 2 7067
− − 781 336 × 0 1386 − − 777 674 × 0 1815 + 58 2911
= 307 779 N m
I y αy + I x − I z ωx ωz = ΣM y = M yp − M yd
M yp = I y αy + I x − I z ωx ωz + M yd
= 0 0024 × − 3 024 + 0 0138 − 0 0138 − 0 0350 × − 2 7067
+ − 103 85 = − 103 857 N m
I z αz + I y − I x ωx ωy = ΣM z = − Rxd ld − Rxp lp + M zp − M zd
M zp = I z αz + I y − I x ωx ωy + Rxd ld + Rxp lp + M zd
= 0 0138 × − 17 076 + 0 0024 − 0 0138 − 0 0350 × − 2 7067
+ − 170 603 × 0 1386 + − 166 26 × 0 1815 + − 13 5723
= − 67 631 N m
The interpretation of these moments for the left knee as the subject bears weight during single
support is as follows. Mxp is +ve; thus, it is a counterclockwise moment and hence an abductor
moment acting at the knee to counter the large gravitational load of the upper body acting down-
ward and medial of the support limb. Mzp is the axial moment acting along the long axis of the leg
and reflects the action of the left hip internal rotators actively rotating the pelvis, upper body, and
right limb in a forward direction to gain extra step length. Myp is −ve, indicating a clockwise
(flexor) knee moment in the sagittal plane that would assist in starting the knee to flex late in
stance shortly before TO.
The next stage of the kinetic analysis is to transform
(which are in the leg anatomical axis system), to the global
ISTUDY
analysis can continue for the thigh segment. This transformation is accomplished by the [A to G]
matrix for the leg and yields a new set of distal moments, MXD, MYD, and MZD, for the thigh
analysis.
As an exercise, students may wish to repeat the preceding calculations for frame 5 or 7.
For frame 5: aX = 5.89 m/s2, aY = 1.30 m/s2, aZ = − 1.66 m/s2; for frame 7: aX = 9.60 m/s2,
aY = 1.94 m/s2, aZ = − 0.020 m/s2.
7.4.4 Joint Mechanical Powers

The joint mechanical power generated or absorbed at the distal and proximal joints, Pd and Pp,
can now be calculated using Equation (5.5) for each of three-moment components and their
respective angular velocities:
Pd = M xd ωxd + M yd ωyd + M zd ωzd (7.9a)

Pp = M xp ωxp + M yp ωyp + M zp ωzp (7.9b)
where the moments are previously defined and ωxd, ωyd, ωzd, ωxp, ωyp, and ωzp are the joint angu-
lar velocities at the distal and proximal ends (not shown in Figure 7.4). The angular velocities are
in rad/s, the moments are in N m, and the powers are in W. If these products were +ve the muscle
group concerned would be generating energy, and if they were −ve, the muscle group would be
absorbing energy.
7.4.5 Induced Acceleration Analysis

Thus far in this chapter, we have focused on calculating the position, velocity, acceleration,
moment, and power that is seen on a joint-by-joint basis. However, it is important to recognize
and remember that there are systemic effects to any movement. Any time a moment is applied to
one joint of the body, it will induce an acceleration to any attached segment, no matter how dis-
tant. The examining of this effect is called Induced Acceleration Analysis. For example, consider
a position where you start with your upper arm at your side and your forearm at 90 in front of
you. When you swing your upper arm forward in flexion (applying a moment at the shoulder),
your forearm must follow the upper arm. Because the forearm has inertial properties, it will want
to stay in its original position. Therefore, unless active muscle contraction is applied, as the upper
arm flexes, the forearm will extend at the elbow. This would be the upper arm-inducing accel-
eration on the forearm. If, in the above example, you suddenly stopped moving the upper arm in
flexion, the inertial properties of the forearm will cause it to continue moving and, especially
when it hits the end of range of motion (fully extended), will act to apply an extension moment
on the upper arm. This would be the forearm-inducing acceleration on the upper arm.
As we can see, any start or stop to one segment has interactions with all the other systemic
segments to which it is attached. While the most obvious effects are seen in adjacent segments,
the movements do affect segments further away in the chain as well, such as an ankle plantar-
flexion moment inducing an acceleration of the head. These interactions can be seen with the
mathematical equations of motion for the segments. In the equations, we can see that the angular
acceleration for a segment is determined by both that segment’s moment, angle, angular velocity,
and moment applied by gravity and the moment, angle, angular velocity, and moment applied by
gravity of the attached segments in the system.
It is important to keep this concept in mind when studying the movements of a body. When
segments are moving at high speeds and with large moments the explanation may be more clearly
explained by the movements induced by other parts of the body than by focusing on single seg-
ments. A great example of this is in a baseball pitch. If you looked at the moment generated by
each segment’s muscles individually, it is not likely to be able to explain the speed and power that
is delivered to a baseball. Only when you look at how the
ISTUDY
each other; can you see how the interactions create a cumulative effect that is greater than its
constituent parts.
7.4.6 Sample Moment and Power Curves

Figure 7.5 presents a set of intersubject averaged 3D joint moments at the ankle, knee, and hip
during one walking stride (Eng and Winter, 1995). HC is at 0% stride and TO is at 60%.
Figure 7.6 shows the 3D power curves for these intersubject averages (Eng and Winter,
1995). The curves are normalized relative to body mass; the moments are reported in
N m/kg and the powers in W/kg. Convention plots extensor moments in the sagittal plane
as positive; in the transverse plane, external rotation moments are positive, and in the frontal
plane, evertor moments are positive. A detailed explanation of the specific function of each of
these moments is beyond the scope of this text; however, a few comments will be made on the
larger or functionally more important moments.
1. The largest moment during walking is seen at the ankle in the sagittal plane. Immediately at
HC, there is a small dorsiflexor moment to lower the foot to the ground; this is followed by a
large increase in plantarflexor moment reaching a peak at about 50% of stride to cause the
ankle to rapidly plantarflex and achieve an upward and forward “push off” of the lower limb
as the subject starts swing at TO (see A2-S power generation burst in Figure 7.6).
2. The knee extensors are active at 8–25% of stride to control knee flexion as the limb accepts
weight (K1-S absorption phase), then the moment reverses to a flexor pattern as a by-
product of the gastrocnemic’s contribution to the increasing ankle plantarflexor moment.
Intersubject ankle moments Intersubject knee moments Intersubject hip moments

0.25 2.00
1.50
Frontal
Frontal
Frontal
Evertor
Abductor
Abductor
0 0
–0.15 –0.40
–0.35
0 20 40 60 80 100 0 20 40 60 80 100 0 20 40 60 80 100
0.150 0.15 0.30
Joint moment (N·m/kg)
Transverse Transverse Transverse

Ext rotation
Ext rotation
Ext rotation
0
0
0
–0.035 –0.15 –0.25
0 20 40 60 80 100 0 20 40 60 80 100 0 20 40 60 80 100
2.00 0.90 2.0
Sagittal Sagittal
Sagittal
Plantarflexor
Extensor
Extensor
0 0
0
–0.50 –0.75 –1.5
0 20 40 60 80 100 0 20 40 60 80 100 0 20 40 60 80 100
% of stride % of stride % of stride
Figure 7.5 Typical 3D moments at the ankle, knee, and hip during one stride of walking (heel contact at 0 and
100%). Profiles are intersubject averages; solid line is the average curve with one standard deviation plotted as a
dotted line. For interpretation of the more important profiles, see the
ISTUDY
Intersubject ankle powers Intersubject knee powers Intersubject hip powers

A1–F H3–F
0.15 0.75
0.35 H2–F
Frontal
K1–F Frontal
Frontal
0 0
Gen
Gen
Gen
0
–0.25 K2–F
–0.30 –1.50 H1–F
0 20 40 60 80 100 0 20 40 60 80 100 0 20 40 60 80 100
0.050 0.075 0.080
Joint power (W/kg)
Joint power (W/kg)
Joint power (W/kg)

Transverse 0 0
0 Transverse
Gen
Gen
Gen
Transverse
K1–T
–0.050 –0.300 –0.350 H1–T
0 20 40 60 80 100 0 20 40 60 80 100 0 20 40 60 80 100
6.0 A2–S 1.2 K0–S 2.5
H1–S H3–S
K2–S Sagittal
Sagittal 0 Sagittal
Gen
Gen
Gen
0
K1–S
0
K4–S
–1.5 A1–S –2.5 K3–S –2.0 H2–S
0 20 40 60 80 100 0 20 40 60 80 100 0 20 40 60 80 100

% of stride % of stride % of stride
Figure 7.6 Mechanical power generation and absorption at the ankle, knee, and hip for the same averaged trials as
in Figure 7.5. Generation power is +ve and absorption power is −ve. For interpretation of selected profiles, see
the text.
Then, just before and after TO, a small knee extensor moment acts to limit the amount of
knee flexion in late stance and early swing (K3-S absorption phase). The final burst of
flexor activity just before HC is to decelerate the swinging leg prior to HC (K4-S absorption
phase).
3. The hip pattern is characterized by an extensor moment for the first half of stance, followed
by a flexor for the latter half. During the first half, the extensors stabilize the posture of the
trunk by preventing it from flexing forward under the influence of a large posterior reaction
force at the hip; this also assists the knee extensors in preventing collapse of the knee joint
and, in addition, contributes to forward propulsion by what has been described as a “push
from behind” (H1-S generation burst). The flexor moment during the second half of stance
serves two functions: first, to stabilize the trunk posture by preventing it from flexing back-
ward under the influence of the forward reaction force at the hip; second, in the last phase of
stance and early swing (50–75% of stride), to achieve a “pull-off” of the thigh into swing
(H3-S generation burst).
4. The major transverse activity is at the hip. During the first half of stance, the external rota-
tors of the stance limb act to decelerate the horizontal rotation of the pelvis (and trunk) over
the stance limb (H1-T absorption phase); then, during the second half, the internal rotators
are active to stabilize the forward rotation of the pelvis and swing limb.
5. The frontal plane moments at the stance hip are a strong abductor pattern to prevent drop
of the pelvis (and entire upper body) against the forces of gravity, which are acting about
10 cm medial of the stance hip (H1-F absorption phase as the pelvis drops, followed by
H2-F and H3-F generation phases as the pelvis, trunk, and swing limb are lifted to help
achieve a safe toe clearance during swing). It is
ISTUDY
at the stance knee, but this is not as a result of muscle activity. Rather, it is the response of
the weight-bearing knee to the large gravitational load; the knee tries to invert but the pas-
sive loading of the medial condyles and unloading of the lateral condyles creates an internal
abductor moment. Such an example illustrates the fact that internal skeletal and ligament
structures can aid (or in some cases hinder) the activity of the muscles. Rehabilitation engi-
neers involved in prothesis design must be aware of the internal moments created by the
springs, dampers, and mechanical stops in their design.
7.5 SUGGESTED FURTHER READING

D’Sousa, A. F. and V. J. Garg. Advanced Dynamics Modeling and Analysis. (Prentice-Hall, Englewood Cliffs,
NJ, 1984).
Greenwood, D. T. Principles of Dynamics, 2nd edition. (Prentice-Hall, Englewood Cliffs, NJ, 1988),
Chapter 7.
Zatsiorsky, V. M. Kinetics of Human Motion. (Human Kinetics, Champaign, IL, 2002).
7.6 REFERENCES
Collins, T. D., S. N. Ghoussayni, D. J. Ewins, and J. A. Kent. “A Six Degrees-of-Freedom Marker Set for Gait
Analysis: Repeatability and Comparison with a Modified Helen Hayes Set,” Gait Posture 30 (2):
173–180, 2009.
Davis, R. B., S. Õunpuu, D. Tyburski, and J. R. Gage. “A Gait Analysis Data Collection and Reduction
Technique,” Hum. Mov. Sci. 10: 575–587, 1991.
Eng, J. J. and D. A. Winter. “Kinetic Analysis of the Lower Limbs During Walking: What Information Can Be
Gained from a Three-dimensional Model,” J. Biomech. 28: 753–758, 1995.
Õunpuu, S., R. B. Davis, and P. A. DeLuca. “Joint Kinetics: Methods, Interpretation and Treatment Decision-
Making in Children with Cerebral Palsy and Myelomeningocele,” Gait Posture 4: 62–78, 1996.
Wu, G., S. Siegler, P. Allard, C. Kirtley, A. Leardini, D. Rosenbaum, M. Whittle, D. D. D’Lima,
L. Cristofolini, W. Harmut, O. Schmid, and I. Stokes. “ISB Recommendation On Definitions of Joint
Coordinate System of Various Joints For The Reporting of Human Joint Motion—Part 1: Ankle, Hip,
and Spine,” J. Biomech. 35: 543–548, 2002.
Wu, G., F. C. T. van der Helm, H. E. J. Veeger, M. Makhsous, P. Van Roy, C. Anglin, J. Nagels, A. R.
Karduna, K. McQuade, X. Wang, F. W. Werner, and B. Buchholz. “ISB Recommendation On Definitions
of Joint Coordinate Systems of Various Joints For The Reporting of Human Joint Motion—Part 2: Shoul-
der, Elbow, Wrist and Hand,” J. Biomech. 38: 981–992, 2005.
Zeni, J. A., Jr., J. G. Richards, and J. S. Higginson. “Two Simple Methods for Determining Gait Events
During Treadmill and Overground Walking Using Kinematic Data,” Gait Posture 27(4): 710–714, 2008.
ISTUDY
8
MUSCLE MECHANICS
1
2
3
8.0 INTRODUCTION
The most intriguing and challenging area of study in biomechanics is probably the muscle itself.
It is the “living” part of the system. The neural control, metabolism, and biomechanical charac-
teristics of muscle are subjects of continuing research. The purpose of this chapter is to report the
state of knowledge with regard to the biophysical characteristics of individual motor units (MUs),
connective tissue, and the total muscle itself. Each of the biophysical characteristics is described
in detail, and it is shown how these characteristics influence the biomechanical function of the
overall muscle.
8.0.1 The Motor Unit

The smallest subunit that can be controlled is called a MU because it is innervated separately by a
motor axon. Neurologically the MU consists of a synaptic junction in the ventral root of the spinal
cord, a motor axon, and a motor end plate in the muscle fibers. Under the control of the MU are as
few as three muscle fibers or as many as 2000, depending on the fineness of the control required
(Feinstein et al., 1955). Muscles of the fingers, face, and eyes have a small number of shorter
fibers in a MU, while the large muscles of the leg have a large number of long fibers in their
MUs. A muscle fiber is about 100 μm in diameter and consists of fibrils about 1 μm in diameter.
Fibrils, in turn, consist of filaments about 100 Å in diameter. Electron micrographs of fibrils show
the basic mechanical structure of the interacting actin and myosin filaments. In the schematic
diagram of Figure 8.1, the darker and wider myosin protein bands are interlaced with the lighter
and smaller actin protein bands. The space between them consists of a cross-bridge structure, and
it is here that the tension is created and the shortening or lengthening takes place. The term con-
tractile element is used to describe the part of the muscle that generates the tension, and it is this
part that shortens and lengthens as positive or negative work is done. The basic length of the
†
Deceased

Companion website:
ISTUDY
172 MUSCLE MECHANICS
I A I
z H z
Sarcomere
length
(1.6–3.6 μm)
Figure 8.1 Basic structure of the muscle contractile element showing the Z lines and sarcomere length. Wider dark
myosin filament interacts across cross-bridges (cross-hatched lines) with the narrower actin filament. Darker and
lighter bands (A, H, and I ) are shown.
myofibril is the distance between the Z lines and is called the sarcomere length. It can vary from
1.5 μm at full shortening to 2.5 μm at resting length to about 4.0 μm at full lengthening.
The structure of the muscle is such that many filaments are in parallel and many sarcomere
elements are in series to make up a single contractile element. Consider a MU of a cross-sectional
area of 0.1 cm2 and a resting length of 10 cm. The number of sarcomere contractile elements
in series would be 10 cm /2.5 μm = 40, 000 and the number of filaments (each with an area of
10−8 cm2) in parallel would be 0.1/10−8 = 107. Thus, the number of contractile elements of
sarcomere length packed into this MU would be 4 × 1011.
The active contractile elements are contained within another fibrous structure of connective
tissue called fascia. These connective tissue coverings enclose each of the muscle’s subscales.
Individual muscle fibers are enclosed by endomysium. A group of fibers or fascicles are then
enclosed by perimysium. Lastly, a group of fascicles or whole muscles are enclosed by epimy-
sium. Ultimately, these connective tissue layers converge at both ends to form tendons. The
mechanical characteristics of connective tissue are important in the overall biomechanics of the
muscle. Some of the connective tissue is in series with the contractile element, while some is
in parallel. The effect of this connective tissue has been modeled as springs and viscous dampers.
8.0.2 Recruitment of Motor Units

Each muscle has a finite number of MUs, each of which is controlled by a separate nerve ending.
Excitation of each unit is an all-or-nothing event. The electrical indication is a MU action poten-
tial; the mechanical result is a twitch of tension. An increase in tension within a single muscle can,
therefore, be accomplished in two ways: by an increase in the stimulation rate for that MU and/or
by the excitation (recruitment) of an additional MU. Figure 8.2 shows the electromyogram
(EMG) of a needle electrode in a muscle as the tension was gradually increased. The upper tracing
shows one MU firing, the middle trace, two MUs, and the lower trace three units. Initially, muscle
tension increases as the firing rate of the first MU increases. Each unit has a maximum firing rate,
and this rate is reached well after the next unit is recruited (Erim et al., 1996). When the tension is
reduced, the reverse process occurs. The firing rates of all recruited units decrease until the last-
recruited unit drops out at a rate usually less than when it was originally recruited. As the tension
further decreases, the remainder of the units drops out in the reverse order from that in which they
were recruited. The firing rate of individual units increases monotonically with force; however,
the collective firing rates of all active MUs increase nonlinearly as force increases from 0 to 100%
maximum voluntary contraction (MVC) (Erim et al.,
ISTUDY
Figure 8.2 EMG from an indwelling electrode in a muscle as it begins to develop tension. The smallest motor unit
is first recruited, and as its rate increases, a second, then a third motor unit are recruited. Each motor action potential
has a characteristic shape at a given electrode, which depends on the size of the motor unit and the distance from the
electrode to the fibers of the motor unit (see Chapter 9).
firing pattern of MUs within a single muscle. This strategy is thought to prevent fatigue of the
larger MUs. Mathematical models describing recruitment have been reported by Wani and Guha
(1975), Milner-Brown and Stein (1975), Fuglevand et al. (1993), and Erim et al. (1996).
8.0.3 Size Principle

Considerable research and controversy have taken place over the past several decades over how
the MUs are recruited. Which ones are recruited first? Are they always recruited in the same
order? It is now generally accepted that they are recruited according to the size principle
(Henneman, 1974a), which states that the size of the newly recruited MU increases with the ten-
sion level at which it is recruited. This means that the smallest unit is recruited first and the largest
unit last. In this manner, low-tension movements can be achieved in finely graded steps. Con-
versely, those movements requiring high forces but not needing fine control are accomplished
by recruiting the larger MUs. Figure 8.3 depicts a hypothetical tension curve resulting from suc-
cessive recruitment of several MUs. The smallest MU (MU 1) is recruited first, usually at an ini-
tial frequency ranging from about 5 to 13 Hz. Tension increases as MU 1 fires more rapidly, until
a certain tension is reached at which MU 2 is recruited. Here, MU 2 starts firing at its initial low
rate, and further tension is achieved by the increased firing of both MU 1 and 2. At a certain
tension, MU 1 will reach its maximum firing rate (15–60 Hz) and will, therefore, be generating
its maximum tension. This process of increasing tension, reaching new thresholds, and recruiting
another larger MU continues until MVC is reached (not shown in Figure 8.3). At that point, all
MUs will be firing at their maximum frequencies but in an asynchronous pattern to prevent
fatigue. Tension is reduced by the reverse process: successive reduction of firing rates and drop-
ping out of the larger units first (DeLuca et al., 1982). During rapid ballistic movements, the firing
rates of individual MUs have been estimated to reach 120 Hz (Desmedt and Godaux, 1978).
The MU action potential increases with the size of the MU with which it is associated (Milner-
Brown and Stein, 1975). The reason for this appears to be twofold. The larger the MU, the larger
the motor neuron that innervates it, and the greater the depolarization potentials seen at the motor
end plate. Second, the greater the mass of the MU, the greater the voltage changes seen at a given
electrode. However, it is never possible from a given
ISTUDY
Max.
Muscle tension tension M.U.4
Threshold
M.U.4 Max. tension M.U.3
Threshold
Threshold
T
M.U.1
M.U.2
M.U.3
M.U.4
Figure 8.3 Size principle of recruitment of motor units. Smaller motor units are recruited first; successively larger
units begin firing at increasing tension levels. In all cases, the newly recruited unit fires at a base frequency then
increases to a maximum.
because its action potential also decreases with the distance between the recording site and the
electrode. Thus, a large MU located a distance from the electrode may produce a smaller action
potential than that produced by a small MU directly beneath the electrode.
8.0.4 Types of Motor Units – Fast- and Slow-Twitch Classification

There have been many criteria and varying terminologies associated with the types of MUs pres-
ent in any muscle (Henneman, 1974b). Biochemists have used metabolic or staining measures to
categorize the fiber types. Biomechanics researchers have used force (twitch) measures (Milner-
Brown et al., 1973b), and electrophysiologists have used EMG indicators (Wormolts and Engel,
1973; Milner-Brown and Stein, 1975). The smaller slow-twitch MUs have been called tonic
units. Histochemically they are the smaller units (type I), and metabolically they have fibers rich
in mitochondria, are highly vascularized, and therefore have a high capacity for aerobic metab-
olism. Mechanically they produce twitches with a low peak tension and a long time to peak
(60–120 ms). The larger fast-twitch MUs are called phasic units (type II). They have less mito-
chondria, are poorly vascularized, and therefore rely primarily on anaerobic metabolism. The
twitches have larger peak tensions in a shorter time (10–50 ms). Figure 8.4 is a typical histochem-
ical stain of fibers of a muscle that contains both slow- and fast-twitch fibers. An adenosine tri-
phosphatase (ATPase)-type stain was used here, so slow-twitch fibers appear dark and fast-twitch
fibers appear light. If an indwelling microelectrode were present in the area of these fibers, the
muscle action potential from these darker slow-twitch fibers would be smaller than that from the
lighter stained fast-twitch fibers.
ISTUDY
Figure 8.4 Histochemical stain showing dark slow-twitch fibers and light fast-twitch fibers. A myofibrillar
ATPase stain, pH 4.3, was used to stain the vastus lateralis of a female volleyball player. (Reproduced by
permission of Professor J. A. Thomson, University of Waterloo, Waterloo, Ont., Canada.)
In spite of the histochemical classification described, there is growing biophysical evidence

(Milner-Brown et al., 1973b) that the MUs controlled by any motor neuron pool form a contin-
uous spectrum of sizes and excitabilities.
8.0.5 The Muscle Twitch

Thus far, we have said very little about an individual MU twitch. As was described in
Section 8.0.4, each MU has its unique time course of tension. Although there are individual dif-
ferences in each newly recruited MU, they all have the same characteristic shape. The time-course
curve follows closely with that of the impulse response of a critically damped second-order sys-
tem (Milner-Brown et al., 1973a). The electrical stimulus of a MU, as indicated by this action
potential, is of short duration and can be considered an impulse. The mechanical response to this
impulse is the much longer duration twitch. The general expression for a second-order critically
damped impulse response is
t −t T
F t = F0 e (8.1)
T
For the curve plotted in Figure 8.5, the twitch time, T, is the time for the tension to reach a
maximum, and F0 is a constant for that given MU. T is the contraction time and is larger for
the slow-twitch fibers than for the fast-twitch MUs, while F0 increases for the larger fast-twitch
units. Muscles tested by Buchthal and Schmalbruch (1970) using submaximal stimulations
showed a wide range of contraction times. Muscles of the upper limbs generally had short
T values compared with the leg muscles. Typical mean values of T and their range were:
Triceps brachii 44.5 ms (16–68 ms)

Biceps brachii 52.0 ms (16–85 ms)
Tibialis anterior 58.0 ms (38–80 ms)
Soleus 74.0 ms (52–100 ms)
Medial gastrocnemius 79.0 ms (40–110 ms)
ISTUDY
0.368Fo
F(t) = Fo t e–t/T
Twitch tension
T
0 T 2T 3T 4T
Time
Figure 8.5 Time course of a muscle twitch, modeled as the impulse response of a second-order critically damped
system. Contraction time T varies with each motor unit type, from about 20 to 100 ms. Effective tension lasts until
about 4 T.
These same researchers found that T increased in all muscles as they were cooled; for example,
the biceps brachialis had a contraction time that increased from 54 ms at 37 C to 124 ms at 23 C.
It is evident that the cause of the delayed build-up of tension is the slower metabolic rates and
increased muscle viscosity seen at lower temperatures.
Many other researchers have tested other muscles using different techniques. Milner-Brown
et al. (1973b) looked at the first dorsal interossei during voluntary contractions and measured T to
average 55 ms (±12 ms), while Desmedt and Godaux (1978) calculated a mean of 60 ms on the
same muscle but used supramaximal stimulations. These latter researchers, using the same tech-
nique, reported T for the tibialis anterior as 84 ms and for the soleus as 100 ms. Bellemare et al.
(1983), also using supramaximal stimulations, reported soleus T = 116 ms (±9 ms) and biceps
T = 66 ms (±9 ms). It appears that the contraction time varies considerably depending on the mus-
cle, the person, and the experimental technique employed. It is important to note that in these
studies participants were asked to maintain a certain level of contraction prior to stimulations
of specific MUs thereby eliminating the time required to take the slack out of the passive con-
nective tissue at rest.
8.0.6 Shape of Graded Contractions

The shape of a voluntary tension curve depends to a certain extent on the shape of the individual
muscle twitches. For example, if we elicit a maximum contraction in a certain muscle, the rate
of increase of tension depends on the individual MUs and how they are recruited. Even if all
MUs were turned on at the same instant and fired at their maximum rate, maximum tension
could never be achieved in less than the average contraction time for that muscle. However,
as described in Figure 8.2 and depicted in Figure 8.3, the recruitment of MUs does not take
place all at once during a voluntary contraction. The smaller slow-twitch units are recruited
first in accordance with the size principle, with the largest units not being recruited until tension
has built up in the smaller units. Thus, it can take several hundred milliseconds to reach max-
imum tension. During voluntary relaxation of the same muscle, the drop in tension is governed
by the shape of the trailing edge of the twitch curve. This delay in the drop of tension is even
more pronounced than the rise. This delay, combined with the delay in dropping out of the MUs
themselves according to the size principle, means that a muscle takes longer to turn off than to
turn on. Typical turn-on times are 200 ms and turn-off times 300 ms, as shown in Figure 8.6,
where maximum rapid turn-on and turn-off are plotted showing the tension curve and the asso-
ciated EMG.
ISTUDY
8.1 FORCE–LENGTH CHARACTERISTICS OF MUSCLES 177
Rectus
femoris
Force
500 ms
Vastus
lateralis
Figure 8.6 Tension build-up and decrease during a rapid maximum voluntary contraction and relaxation. The time
to peak tension can be 200 ms or longer, mainly because of the recruitment according to the size principle and
because of delay between each motor unit action potential and twitch tension. Note the presence of tension for
about 150 ms after the cessation of EMG activity.
8.1 FORCE–LENGTH CHARACTERISTICS OF MUSCLES
As indicated in Section 8.0.1, the muscle consists of an active element, called the contractile ele-
ment, and passive connective tissue. The net force–length characteristics of a muscle are a com-
bination of the force–length characteristics of both active and passive elements.
8.1.1 Force–Length Curve of the Contractile Element

The key to the shape of the force–length curve is the changes of the structure of the myofibril at
the sarcomere level (Gordon et al., 1966). Figure 8.7 is a schematic representation of the myosin
z
z
z
z
z
% max. isometric tension
z z z z z
Io
Length of contractile
element
Figure 8.7 Tension produced by a muscle as it changes length about its resting length, l0. A drop in tension on
either side of maximum can be explained by the interactions of
ISTUDY
and actin cross-bridge arrangement. At resting length, about 2.5 μm, there are a maximum num-
ber of cross-bridges between the filaments and, therefore, a maximum tension is possible. As the
muscle lengthens, the filaments are pulled apart, the number of cross-bridges reduces, and ulti-
mate force decreases. At full length, about 4.0 μm, there are no cross-bridges and the force
reduces to zero. As the muscle shortens to less than resting length, there is an overlapping of
the cross-bridges and interference takes place. This results in a reduction of force that continues
until a full overlap occurs, at about 1.5 μm. The force does not drop to zero but is drastically
reduced by these interfering elements. It is important to note that these muscle lengths are not
referencing joint positions instead the lengths of muscle if they were excised from the body. Each
muscle crossing a joint in the body will have varying length positions for a given joint position
and therefore will have altering force-producing capabilities.
8.1.2 Influence of Parallel Connective Tissue

The connective tissue that surrounds the contractile element influences the force–length curve. It
is called the parallel elastic component, and it acts much like an elastic band. When the muscle is
at resting length or less, the parallel elastic component is in a slack state with no tension. As the
muscle lengthens beyond the resting length, the parallel element is no longer loose, so tension
begins to build up, slowly at first and then more rapidly. Unlike most springs, which have a linear
force–length relationship, the parallel element is quite nonlinear. In Figure 8.8, we see the force–
length curve of this element, Fp, combined with that of the overall contractile component Fc. If we
sum the forces from both elements, we see the overall muscle force–length characteristic, Ft. The
force–length curve typically presented is usually for a maximum contraction. The passive force,
Fp, of the parallel element is always present, but the amount of active tension in the contractile
element at any given length is under voluntary control. Thus, the overall force–length character-
istics are a function of the percentage of excitation, as seen in Figure 8.9.
8.1.3 Series Elastic Tissue

All connective tissue in series with the contractile component, including the tendon, is called the
series elastic element. Under isometric contractions, it will lengthen slightly as tension increases.
Fc
Ft
FP
I
Force
Ft
Fc FP
Io
Length
Figure 8.8 Contractile element producing maximum tension Fc along with the tension from Fp from the parallel
elastic element. Tendon tension is Ft = Fc + Fp.
ISTUDY
8.1 FORCE–LENGTH CHARACTERISTICS OF MUSCLES 179
100%
Force
75%
50%
25%
Io
Length
Figure 8.9 Tendon tension resulting from various levels of muscle activation. Parallel elastic element generates
tension independent of the activation of the contractile element.
However, during dynamic situations, the series elastic element, in conjunction with viscous com-
ponents, does influence the time course of the muscle tension.
During isometric contractions, the series elastic component is under tension and, therefore, is
stretched a finite amount. Because the overall length of the muscle is kept constant, the stretching
of the series elastic element can only occur if there is an equal shortening of the contractile ele-
ment itself. This is described as internal shortening. Figure 8.10 illustrates this point at several
contraction levels. Although the external muscle length L is kept constant, the increased tension
from the contractile element causes the series elastic element to lengthen by the same amount as
the contractile element shortens internally. The amount of internal shortening from rest to max-
imum tension is only a few percent of the resting length in most muscles (van Ingen Schenau,
1984) but has been shown to be as high as 7% in others (Bahler, 1967). It is widely inferred that
these series elastic elements store large amounts of energy as muscles are stretched prior to an
explosive shortening in athletic movements. However, van Ingen Schenau (1984) has shown that
the elastic capacity of these series elements is far too small to explain improved performances
resulting from pre-stretch. It is possible that a combination of the stored elastic energy in the
series elastic components, enhanced sensitivity of muscle spindles, and desensitivity of Golgi
tendon organs result in the improved performance seen following a pre-stretch. This can be
achieved through plyometric training programs.
Experiments to determine the force–length characteristics of the series element can only
be done on isolated muscle and will require dynamic changes of force or length. A typical
experimental setup is shown in Figure 8.11. The muscle is stimulated to a certain level of tension
while held at a certain isometric length. The load (tension) is suddenly dropped to zero by
releasing one end of the muscle. With the force suddenly removed, the series elastic element,
which is considered to have no mass, suddenly shortens to its relaxed length. This sudden
shortening can be recorded, and the experiment repeated for another force level until a force-
shortening curve can be plotted. During the rapid shortening period (about 2 ms), it is assumed
that the contractile element has not had time to change length, even though it is still generating
tension. Two other precautions must be observed when conducting this experiment. First, the
isometric length prior to the release must be such that there is no tension in the parallel elastic
element. This will be so if the muscle is at resting length or shorter. Second, the system that
records the sudden shortening of the series elastic
ISTUDY
(L–x0) x0
F = 0 Muscle
relaxed
L
(L–x1) x1
F1 Moderate
S. E.
contraction
P. E.
(L–x2) x2
F2
Heavy
contraction
Figure 8.10 Introduction of the series elastic (S.E.) element. During isometric contractions, the tendon tension
reflects a lengthening of the series element and an internal shortening of the contractile element. During most
human movement, the presence of the series elastic element is not too significant, but during high-performance
movements, such as jumping, it is responsible for storage of energy as a muscle lengthens immediately prior to
rapid shortening.
F
Force
Force
transducer
Release
x
mechanism
Displacement
+ –
x
0 10 20 30 40
Time (ms)
Figure 8.11 Experimental arrangement to determine the spring constant of the series elastic element. Muscle is
stimulated; after the tension builds up in the muscle, the release mechanism activates, allowing an almost
instantaneous shortening, x, while the force change is measured on a force transducer.
viscosity. Because of the high acceleration and velocity associated with the rapid shortening, the
resisting force of the attached recording transducer should be negligible. If such conditions are
not possible, a correction should be made to account for the mass or viscosity of the transducer.
8.1.4 In Vivo Force–Length Measures

Length–tension relationships of striated muscle have been well researched in mammalian and
amphibian species, but in humans, the studies have been limited to changes in joint moments
during MVC as the joint angle is altered. Two problems arise in such studies. First, it is usually
impossible to generate an MVC for a single agonist without activating the remaining agonists.
Thus, the joint moment is the sum of the moments generated by several agonists, each of which is
ISTUDY
8.2 FORCE–VELOCITY CHARACTERISTICS 181
likely to be operating at a different point in its force–length curve. Also, the moment generated by
any muscle is a product of its force and its moment arm length, both of which change as the joint
angle changes. Thus, any change in joint moment with joint angle cannot be attributed uniquely
to the length–tension characteristics of one specific muscle.
One in vivo study that did yield some meaningful results was reported by Sale et al. (1982),
who quantified the plantarflexor moment changes as the ankle angle was positioned over a range
of 50 (30 plantarflexion to 20 dorsiflexion). The contribution of the gastrocnemius was
reduced by having the knee flexed to 90 , thus causing it to be slack. Thus, the contribution
of the soleus was considered to be dominant. Also, the range of movement of the ankle was con-
sidered to have limited influence on the moment arm length of the soleus. Thus, the moment–
ankle angle curve was considered to reflect the force–length characteristics of the soleus muscle.
The results of the MVC, peak twitch tensions, and a 10-Hz tetanic stimulation were almost the
same: the peak tension was seen at 15 dorsiflexion and dropped off linearly to near zero as the
ankle plantarflexed to 30 . Such a curve agreed qualitatively with a muscle whose resting length
was at 15 dorsiflexion and whose cross-bridges reached a full overlap of 30 plantarflexion.
8.2 FORCE–VELOCITY CHARACTERISTICS
The previous section was concerned primarily with isometric contractions, and most physio-
logical experiments are conducted under such conditions. However, movement cannot be
accomplished without a change in muscle length. Alternate shortening and lengthening occur
regularly during any given movement and the velocity associated with the shortening and
lengthening will relate to the overall muscle tension. In the next section, we will discuss the
velocity-based mechanisms to muscle tension.
8.2.1 Concentric Contractions

The tension in a muscle decreases as its shortening velocity increases. The characteristic curve
that describes this effect is called a force–velocity curve and is shown in Figure 8.12. The usual
curve is plotted for a maximum (100%) contraction. However, this condition is rarely seen except
in athletic events, and then only for short bursts of time. In Figure 8.12 the curves for 75, 50, and
25% contractions are shown as well. Isometric contractions lie along the zero-velocity axis of this
graph and should be considered as nothing more than a special case within the whole range of
possible velocities. It should be noted that this curve represents the characteristics at a certain
muscle length. To incorporate length as a variable as well as velocity requires a three-dimensional
plot, as is discussed in Section 8.2.3.
The decrease of tension as the shortening velocity increases has been attributed to two
main causes. A major reason appears to be the loss in tension as the cross-bridges in the con-
tractile element break and then reform in a shortened condition. A second cause appears to be
the fluid viscosity in both the contractile element and the connective tissue. Such viscosity
results in friction that requires an internal force to overcome and, therefore, results in a
reduced tendon force. Whatever the mechanism of lost tension, it is clear that the total effect
is similar to that of viscous friction in a mechanical system and can, therefore, be modeled as
some form of a fluid damper. The loss of tension related to a combination of the number of
cross-bridges breaking and reforming and passive viscosity makes the force–velocity curve
somewhat complicated to describe. If all the viscosity were passive, the slope of the
force–velocity curve would be independent of activation. Conversely, if all the viscosity were
related to the number of active cross-bridges, the slope would be proportional to the activa-
tion. Green (1969) has analyzed families of force–velocity curves to determine the relative
contribution of each of these mechanisms.
ISTUDY
100%
Force
75%
Fmax
M
ax
50%
im
um
te
n
sio
25%
n
– Lengthen 0 Shorten +
Velocity
Figure 8.12 Force–velocity characteristics of skeletal muscle for different levels of muscle activation; shown are
25, 50, 75, and 100% levels of activation. All such measures must be taken as the muscle shortens or lengthens at a
given length, and the length must be reported. During shortening, the curves follow the hyperbolic Hill model, but
during lengthening, the curves depend on the experimental protocol; the solid lines are for isotonic activity, while the
dashed lines are for isovelocity activity.
A curve fit of the force–velocity curve was demonstrated by Fenn and Marsh (1935), who used
the equation:
V = V 0 eP B − KP (8.2)
where
V = shortening velocity at any force
V0 = shortening velocity of unloaded muscle
P = force
B, K = constants.
A few years later, Hill (1938) proposed a different mathematical relationship that bore some
meaning with regard to the internal thermodynamics. Hill’s curve was in the form of a hyperbola,
P + a V + b = P0 + a b (8.3)
where
P0 = maximum isometric tension
a = coefficient of shortening heat
b = a V0/P0
V0 = maximum velocity (when P = 0).
More recently, this hyperbolic form has been found to fit the empirical constant only during
isotonic contractions near resting length.
The maximum velocity of shortening is often expressed as a function of l0, the resting fiber
length of the contractile elements of the muscle. Animal experiments have shown the maximum
shortening velocity to be about 6l0/s (Close, 1965; Faulkner et al., 1980). However, it appears that
this must be low for some human activity. Woittiez (1984) has calculated the shortening velocity
ISTUDY
in soleus, for example, to be above 10l0/s (based on plantarflexor velocity in excess of 8 rad/s and
a moment arm length of 5 cm).
8.2.2 Eccentric Contractions

The vast majority of research done on isolated muscle during in vivo experiments has involved
concentric contractions. As a result, there is relatively limited knowledge about the details of the
force–velocity curve as the muscle lengthens. The curve certainly does not follow the detailed
mathematical relationships that have been developed for concentric contractions.
This lack of information about eccentric contractions is unfortunate because normal human
movement usually involves as much eccentric as concentric activity. If we neglect air and ground
friction, level walking involves equal amounts of positive and negative work, and in downhill
gait, negative work dominates. Figure 8.12 shows the general shape of the force–velocity curve
during eccentric contractions. It can be seen that this curve is an extension of the concentric curve.
During isotonic eccentric action, the solid line curves apply (Winters, 1990), but during isove-
locity eccentric activity, the relationship follows the dotted lines shown in Figure 8.12 (Zahalak,
1990; Sutarno and McGill, 1995). If the maximum isometric force is Fmax, the plateau reached in
the eccentric phase varies from 1.1 Fmax to 1.8 Fmax (Winters, 1990). The reason was given for the
force increasing as the lengthening velocity increases was that within the cross-bridges the force
required to reverse pivot the myosin heads and break the bonds is greater than that required to
hold it at its isometric length. The plateau is reached at higher velocities when the cross-bridge
links simply “give way” to produce no further increase in force. Also, there is viscous friction in
the fluid surrounding the muscle fibers, and this friction force must be overcome as well. This is
similar to opening a screen door (with a pneumatic door closer) very quickly. Air pressure is built-
up within the pneumatic closer since air cannot escape at that rate through the small hole. During
eccentric muscle contractions, the connective tissue element will undergo tension and water will
be forced out of the tissue. When lengthened quickly the water cannot escape at that rate and
water pressure will increase, creating viscous friction. However, the distributed moment (DM)
state variable model that approximates the Huxley-type cross-bridge theories (Zahalak and
Ma, 1990) predicted the drop in tension reported in the eccentric region for constant velocity
stretch.
Experimentally, it is somewhat more difficult to conduct experiments involving eccentric
work because an external device must be available to do the work on the human muscle. Such
a requirement means that a motor is needed to provide an external force that will always exceed
that of the muscle. Experiments on isolated muscle are safe to conduct, but in vivo experiments on
humans are difficult because such a machine could cause lengthening even past the safe range of
movement of a joint. The excessive force could tear the limb apart at the joint. Foolproof safety
mechanisms have to be installed to prevent such an occurrence. In addition to safety concerns, it
needs to be highlighted that joints have several muscles combining to create the internal moment
which will counteract the external load. Therefore, it is impossible to isolate muscles during such
experiments as these devices can only apply external force to a joint.
8.2.3 Combination of Length and Velocity Versus Force

In Section 8.1 and in this section, it is evident that the tendon force is a function of both length and
velocity. Therefore, a proper representation of both these effects requires a three-dimensional plot
like that shown in Figure 8.13. The resultant curve is actually a surface, which here is represented
only for the maximum contraction condition. The more normal contractions are at a fraction of
this maximum, so that surface plots would be required for each level of muscle activation, say at
75, 50, and 25%.
ISTUDY
1.2 1.4
th–
Force (fraction of Po)

Leng n For
ce–
n s io ve
te
e plan locity
plan e
.8
.6
.4
1.3
1.2
.2
1.1
) Vel
ocit
.9 o n at Lo y–
.8 ra c t i p lane length
Ne
g
.7
e n g th (f .5 Vel
ocit
wo ative L y (L
rk 1.0 o /s)
Pos 1.5
itive 2.0
wor
k
Figure 8.13 Three-dimensional plot showing the change in contractile element tension as a function of both
velocity and length. Surface shown is for maximum muscle activation; a new “surface” will be needed to
describe each level of activation. Influence of parallel elastic element is not shown.
8.2.4 Combining Muscle Characteristics with Load Characteristics: Equilibrium

Muscles are the only motors in the human system, and when they are active, they must be in
equilibrium with their load. The load can be static, for example, when holding a weight against
gravity or applying a static force against a fixed object (e.g. a wall, the floor), or in an isometric
co-contraction, where one muscle provides the load of the other. Or the load may be dynamic and
the muscles are accelerating or decelerating an inertial load or overcoming the friction of a vis-
cous load. In the vast majority of voluntary movements, there is a mixture of static and dynamic
load. The one condition that is satisfied during the entire movement is that of equilibrium: at any
given point in time, the operating point will be the intersection of the muscle and load character-
istics. In a dynamic movement, this operating or equilibrium point will be continuously changing.
In Chapter 5, an equilibrium was recognized every time the equations of motion were written.
However, such equations recognized the net effect of all muscles acting at each joint and
represented them as a torque motor without any concern about the internal characteristics of
the muscles themselves. These are now examined.
Consider a muscle that is contracting against a spring-like load that can have a linear or a non-
linear force–length characteristic. Figure 8.14a plots the force–length characteristics of a muscle
at four different levels of activation, along with the linear and nonlinear characteristics of the
spring loads. Assume that the muscle is at a length l1 longer than the resting length, l0 when
the spring is at rest. The linear spring is compressed with a 50% muscle contraction, the equi-
librium point a is reached, and the muscle will shorten from l1 to l2. When the nonlinear spring
is compressed with a 50% contraction, the muscle shortens from l1 to l3 and the equilibrium point
b is reached. If a gravitational load is lifted by the elbow flexors, the flexor muscles have the
characteristics shown in Figure 8.14b. Starting with the forearm lowered to the side so that it
is vertical, the initial equilibrium point is a. Then, as the flexors contract and shorten, equilibrium
points b through e are reached. Finally, as the muscles are activated 100%, the equilibrium point,
e, is reached. During the transient conditions en route from
ISTUDY
(a)
100%
Force
75%
50%
b
a 25%
I2 I3 I1
I0
Length
(b)
100%
Force
75%
50%
ed
c
25%
b
a
I0
Length
Figure 8.14 (a) Force–length characteristics of a muscle at four different levels of activation along with the
characteristics of a linear and nonlinear spring. Intersection of spring (load) and muscle characteristics is the
equilibrium point. (b) Equilibrium points of a load and elbow flexor muscle as the elbow is flexed against a
gravitational load.
move about on the force–velocity characteristics, tracing a three-dimensional locus. The concept
of dynamic equilibrium is now addressed.
We use the data from a typical walking stride to plot the time course of the contractions of the
muscles about the ankle. Because the angular changes of the ankle are relatively small, we can
consider that the muscle lengths are proportional to the ankle angle, and these length changes are
also small. Thus, we can plot the event on a two-dimensional force–velocity curve. The length-
ening or shortening velocity can be considered proportional to the angular velocity, and the ten-
don forces are considered proportional to the muscle moment. Thus, using in vivo data, the
traditional force–velocity curve becomes a moment–angular velocity curve. Figure 8.15 is the
resultant plot for the period of time from heel contact to toe-off.
It can be seen from this time course of moment and angular velocity that this common move-
ment is, in fact, quite complex. Contrary to what might be implied by force–velocity curves, a
muscle does not operate along any simple curve, but actually goes through a complex combina-
tion of force and velocity changes. Initially, the dorsiflexor muscles are active as the foot plantar-
flexes between the time of heel contact and flat foot (when the ground reaction force lies behind
the ankle joint). Negative work is being done by the dorsiflexors as they lower the foot to the
ground. After flat foot, the plantarflexors dominate and
ISTUDY
140
Pla
nta
120 rm
om
Net ankle moment (N·m)
40
en
100 %
t
co
nt
r. o
fp
lan
80 tar
Hee
off
60
20%
con
tr. o
plan f
40 tar.
Dors
ifl. m
om
en Heel
20 t contact
Foot flat Toe off
–2 –1 0 1 2 3 4
Dorsiflexion Angular velocity (rad/s) Plantarflexion
Figure 8.15 Time history of muscle moment–velocity during stance phase of walking. Dorsiflexor moment is
shown as a dashed line; plantarflexor moment is a solid line. Stance begins with negative work, then alternating
positive and negative, and finally, a major positive work burst later in push-off, ending at toe-off.
the leg as it rotates over the foot, which is now fixed on the floor. Again, this is negative work.
A few frames after heel-off, positive work begins, as indicated by the simultaneous plantarflexor
muscle moment and plantarflexor velocity. This period is the active push-off phase, when most of
the “new” energy is put back into the body.
8.3 TECHNIQUE TO MEASURE IN VIVO TENDON MECHANICAL

PROPERTIES
Methodologies have been developed and tested to calculate tendon stiffness and force during
isometric contractions. The majority of the literature is focused on the Achilles tendon; however,
similar techniques have been performed at other tendons, such as the patellar tendon. These tech-
niques are useful to examine the effectiveness of surgical and rehabilitation approaches. The fol-
lowing section will describe the detailed steps to accurately calculate Achilles tendon stiffness
and force.
8.3.1 Ankle Joint Moment

The first step in determining Achilles tendon stiffness is to measure ankle joint moments while
performing a series of targeted isometric contractions based on a maximal voluntary isometric
contraction (MVIC) using a dynamometer. Participants lie prone on the dynamometer table with
their knee in 0 of flexion, and their ankle in 20 of plantarflexion. This lower extremity position
is important to minimize the passive torque from the connective tissue structures. While other
lower extremity positions can also minimize the passive connective tissue torque, it is important
to keep the position standardized across participants. Care should be taken to align the ankle axis
of rotation with the dynamometer’s axis of rotation. Even with proper alignment and fixation, the
axis of rotation of the ankle will shift during an isometric
the calculation of the joint moment and tendon elongation.
ISTUDY
8.4 REFERENCES 187
ankle and the dynamometer attachment need to be included to correct joint moments and tendon
elongations (Magnusson et al., 2001). In addition, EMG is often used to correct for co-activation
of the tibialis anterior muscle during a plantarflexion isometric contraction (Maganaris et al.,
1998). During testing, it is necessary to collect passive plantar and dorsiflexion trials to correct
for gravitational forces. In addition, a maximal dorsiflexion contraction is performed to later esti-
mate the ankle dorsiflexion moment (antagonist co-contraction) occurring during a plantarflexion
contraction (Arampatzis et al., 2005).
8.3.2 Tendon Mechanical Properties

Diagnostic ultrasound is used to measure the displacement of the myotendinous junction (MTJ)
of the medial gastrocnemius muscle during isometric plantarflexion. To account for MTJ dis-
placement resulting from joint rotation during the isometric contractions, ultrasound should also
be used during the passive plantarflexion and dorsiflexion trials. To correct the displacement, a
simple difference was taken between active and passive trials. Next, the resting length of the
Achilles tendon needs to be measured in 20 of plantarflexion to minimize any torque to the con-
nective tissue structures. The resting length is measured from the MTJ along the curved path to
the insertion on the calcaneal tuberosity (Arampatzis et al., 2008). Achilles tendon moment arms
are typically estimated based on the relationship between MTJ elongation and ankle joint rotation
(Fath et al., 2010). To calculate tendon force, the ankle joint moment is divided by the ankle joint
moment arms. Tendon stiffness will then be calculated as the slope of the tendon force/elongation
curve. This curve will typically be created using a series of targeted isometric contraction levels
based on the MVIC.
Calculating tendon properties in vivo can be very useful to determine the effectiveness of inter-
ventions designed to improve healing following surgery or rehabilitation protocols. Although the
most established methods have been developed and tested at the Achilles tendon, other tendons
are possible. When attempting these calculations at other joints, it is important to consider
the previously described variables (gravitational forces, moment arm changes, axis of rotation
misalignment, joint rotation, etc.) in addition to isolating the contribution of a single muscle/
tendon unit.
8.4 REFERENCES
Arampatzis, A., S. Stafilidis, G. DeMonte, K. Karamanidis, G. Morey-Klapsing, and G. P. Brüggemann.
“Strain and Elongation of the Human Gastrocnemius Tendon and Aponeurosis During Maximal Plantar-
flexion Effort,” J. Biomech. 38(4): 833–841, 2005.
Arampatzis, A., G. De Monte, and K. Karamanidis. “Effect of Joint Rotation Correction When Measuring
Elongation of the Gastrocnemius Medialis Tendon and Aponeurosis,” J. Electromyogr. Kinesiol.
18(3): 503–508, 2008.
Bahler, A. S. “Series Elastic Component of Mammalian Muscle,” Am. J. Physiol. 213: 1560–1564, 1967.
Bellemare, F., J. J. Woods, R. Johansson, and B. Bigland-Ritchie. “Motor Unit Discharge Rates in Maximal
Voluntary Contractions in Three Human Muscles,” J. Neurophysiol. 50: 1380–1392, 1983.
Buchthal, F. and H. Schmalbruch. “Contraction Times and Fibre Types in Intact Human Muscle,” Acta Phy-
siol. Scand. 79: 435–452, 1970.
Close, R. “The Relation Between Intrinsic Speed of Shortening and Duration of the Active State of Muscles,”
J. Physiol. 180: 542–559, 1965.
DeLuca, C. J., R. S. LeFever, M. P. McCue, and A. P. Xenakis. “Control Scheme Governing Concurrently
Active Motor Units During Voluntary Contractions,” J. Physiol. 329: 129–142, 1982.
Desmedt, J. E. and E. Godaux. “Ballistic Contractions in Fast and Slow Human Muscles: Discharge Patterns
of Single Motor Units,” J. Physiol. 285: 185–196, 1978.
Erim, Z., C. J. DeLuca, and K. Mineo. “Rank-Ordered Regulation
573, 1996.
ISTUDY
Fath, F., A. J. Blazevich, C. M. Waugh, S. C. Miller, and T. Korff. “Direct Comparison of in vivo Achilles Tendon
Moment Arms Obtained from Ultrasound and MR Scans,” J. Appl. Physiol. 109(6): 1644–1652, 2010.
Faulkner, J. A., J. H. Niemeyer, L. C. Maxwell, and T. P. White. “Contractile Properties of Transplanted
Extensor Digitorum Longus Muscle of Cats,” Am. J. Physiol. 238: 120–126, 1980.
Feinstein, B., B. Lindegard, E. Nyman, and G. Wohlfart. “Morphological Studies of Motor Units in Normal
Human Muscles,” Acta Anat. 23: 127–142, 1955.
Fenn, W. O. and B. S. Marsh. “Muscular Force at Different Speeds of Shortening,” J. Physiol. Lond.
85: 277–297, 1935.
Fuglevand, A. J., D. A. Winter, and A. E. Patla. “Models of Recruitment and Rate Coding Organization in
Motor Unit Pools,” J. Neurophysiol. 70: 2470–2488, 1993.
Gordon, A. M., A. F. Huxley, and F. J. Julian. “The Variation is Isometric Tension with Sarcomere Length in
Vertebrate Muscle Fibres,” J. Physiol. 184: 170, 1966.
Green, D. G. “A Note on Modelling in Physiological Regulators,” Med. Biol. Eng. 7: 41–47, 1969.
Henneman, E. “Organization of the Spinal Cord,” in Medical Physiology, 13th edition, Vol. 1, V. B. Mount-
castle, Ed. (C. V. Mosby, St. Louis, MO, 1974a).
Henneman, E. “Peripheral Mechanism Involved in the Control of Muscle,” in Medical Physiology, 13th edi-
tion, Vol. 1, V. B. Montcastle, Ed. (C. V. Mosby, St. Louis, MO, 1974b).
Hill, A. V. “The Heat of Shortening and Dynamic Constants of Muscle,” Proc. Roy. Soc. B 126:
136–195, 1938.
van Ingen Schenau, G. J. “An Alternate View of the Concept of Utilization of Elastic Energy in Human
Movement,” Human Movement Sci. 3: 301–336, 1984.
Maganaris, C. N., V. Baltzopoulos, and A. J. Sargeant. “Differences in Human Antagonistic Ankle Dorsi-
flexor Coactivation between Legs; Can They Explain the Moment Deficit in the Weaker Plantarflexor
Leg?” Exp. Physiol. 83 (6): 843–855, 1998.
Magnusson, S. P., P. Aagaard, P. Dyhre-Poulsen, and M. Kjaer. “Load-Displacement Properties of the Human
Triceps Surae Aponeurosis In Vivo,” J. Physiol. 531(Pt 1): 277–288, 2001.
Milner-Brown, H. S. and R. B. Stein. “The Relation Between the Surface Electromyogram and Muscular
Force,” J. Physiol. 246: 549–569, 1975.
Milner-Brown, H. S., R. B. Stein, and R. Yemm. “The Contractile Properties of Human Motor Units During
Voluntary Isometric Contractions,” J. Physiol. 228: 285–306, 1973a.
Milner-Brown, H. S., R. B. Stein, and R. Yemm. “The Orderly Recruitment of Human Motor Units During
Voluntary Isometric Contractions,” J. Physiol. 230: 359–370, 1973b.
Sale, D., J. Quinlan, E. Marsh, A. J. McComas, and A. Y. Belanger. “Influence of Joint Position on Ankle
Plantarflexion in Humans,” J. Appl. Physiol. 52: 1636–1642, 1982.
Sutarno, C. G. and S. M. McGill. “Isovelocity Lengthening Behaviour of Erector Spinae Muscles,” Eur. J.
Appl. Physiol. 70: 146–153, 1995.
Wani, A. M. and S. K. Guha. “A Model for Gradation of Tension Recruitment and Rate Coding,” Med. Biol.
Eng. 13: 870–875, 1975.
Winters, J. M. “Hill-Based Muscle Models: A System Engineering Perspective,” in Multiple Muscle Systems:
Biomechanics and Movement Organization, J. M. Winters and S. L. J. Woo, Eds. (Springer, New York,
1990), pp. 69–93.
Woittiez, R. O. “A Quantitative Study of Muscle Architecture and Muscle Function,” Ph.D. dissertation, Free
University (Amsterdam), The Netherlands, 1984.
Wormolts, J. R. and W. K. Engel. “Correlation of Motor Unit Behaviour with Histochemical-Myofiber Type
in Humans by Open-Biopsy Electromyography,” in New Developments in Electromyography and Clinical
Neurophysiology, Vol. 1, J. E. Desmedt, Ed. (Karger, Basel, Switzerland, 1973).
Zahalak, G. I. “Modelling Muscle Mechanics (and Energetics),” in Multiple Muscle Systems: Biomechanics
and Movement Organization, J. M. Winters and S. L. J. Woo, Eds. (Springer, New York, 1990), pp. 1–23.
Zahalak, G. I. and S.-P. Ma. “Muscle Activation and Contraction: Constitutive Relations Based Directly on
Cross-Bridge Kinetics,” J. Biomech. Eng. 112: 52–62, 1990.
ISTUDY
9
KINESIOLOGICAL ELECTROMYOGRAPHY
1
2
3
9.0 INTRODUCTION
The electrical signal associated with the contraction of a muscle is called an electromyogram, or
EMG. The study of EMGs, called electromyography, has revealed some basic information; how-
ever, much remains to be learned. Voluntary muscular activity results in an EMG that increases in
magnitude with the tension. However, there are many variables that can influence the signal at
any given time: velocity of shortening or lengthening of the muscle, rate of tension build-up,
fatigue, and reflex activity. An understanding of the electrophysiology and the technology of
recording is essential to the appreciation of the biomechanical relationships that follow.
9.1 ELECTROPHYSIOLOGY OF MUSCLE CONTRACTION
It is important to realize that muscle tissue conducts electrical potentials somewhat similarly to
the way axons transmit action potentials. The name given to this special electrical signal gener-
ated in the muscle fibers as a result of the recruitment of a motor unit is a motor unit action poten-
tial (m.u.a.p.). Electrodes placed on the surface of a muscle or inside the muscle tissue
(indwelling electrodes) will record the algebraic sum of all m.u.a.p.’s being transmitted along
the muscle fibers at that point in time. Those motor units far away from the electrode site will
result in a smaller m.u.a.p. than those of similar size near the electrode.
9.1.1 Motor End Plate

For a given muscle, there can be a variable number of motor units, each controlled by a motor
neuron through special synaptic junctions called motor end plates. An action potential transmit-
ted down the motoneuron (sometimes called the final common pathway) arrives at the motor end
†
Deceased

Companion website:
ISTUDY
190 KINESIOLOGICAL ELECTROMYOGRAPHY
plate and triggers a sequence of electrochemical events. A quantum of Acetylcholine (ACh) is

released; it crosses the synaptic gap (200–500 Å wide) and causes a depolarization of the post-
synaptic membrane. Such a depolarization can be recorded by a suitable microelectrode and is
called an end plate potential (EPP). In normal circumstances, the EPP is large enough to reach a
threshold level, and an action potential is initiated in the adjacent muscle fiber membrane. In dis-
orders of neuromuscular transmission (e.g. depletion of ACh), there may not be a one-to-one rela-
tionship between each motor nerve action potential and a m.u.a.p. The end plate block may be
complete, or it may occur only at high stimulation rates or intermittently.
9.1.2 Sequence of Chemical Events Leading to a Twitch

The beginning of the m.u.a.p. starts at the Z line of the contractile element (see Chapter 8) by
means of an inward spread of the stimulus along the transverse tubular system. This results in
a release of Ca2 + in the sarcoplasmic reticulum. Ca2 + rapidly diffuses to the contractile filaments
of actin and myosin, where adenosine triphosphate (ATP) is hydrolyzed to produce adenosine
diphosphate (ADP) plus heat plus mechanical energy (tension). The mechanical energy manifests
itself as an impulsive force at the cross-bridges of the contractile element. The time course of the
contractile element’s force has been the subject of considerable speculation and has been modeled
mathematically, as described in Chapter 8.
9.1.3 Generation of a Muscle Action Potential

The depolarization of the transverse tubular system and the sarcoplasmic reticulum results in a
depolarization “wave” along the direction of the muscle fibers. It is this depolarization wave front
and the subsequent repolarization wave that are “seen” by the recording electrodes.
Many types of EMG electrodes have developed over the years, but generally, they can be
divided into two groups: surface and indwelling (intramuscular). The seminal paper by Basma-
jian (1973) gives a detailed review of the use of different types along with their connectors. Sur-
face electrodes consist of disks of metal, usually silver/silver chloride, of about 1 cm in diameter.
These electrodes detect the average activity of superficial muscles and give more reproducible
results than do indwelling types (Komi and Buskirk, 1970; Kadaba et al., 1985). Smaller disks
can be used for smaller muscles. Indwelling electrodes are required, however, for the assessment
of fine movements or to record from deep muscles. A needle electrode is a fine hypodermic nee-
dle with an insulated conductor located inside and bared to the muscle tissue at the open end of the
needle; the needle itself forms the other conductor (Figure 9.1). For research purposes, multielec-
trode types have been developed to investigate the “territory” of a motor unit (Buchthal et al.,
1959), which has been found to vary from 2 to 15 mm in diameter. Fine-wire electrodes, which
Figure 9.1 Intramuscular EMG electrode with fine wire passing

hooked ends for fixation within muscle tissue.
ISTUDY
9.1 ELECTROPHYSIOLOGY OF MUSCLE CONTRACTION 191
have a diameter about that of human hairs, are now widely used. They require a hypodermic nee-
dle to insert. After removal of the needle, the fine wires with their uninsulated tips remain inside
in contact with the muscle tissue. A comparison of this experimental investigation of motor unit
territory was seen to agree with theoretical predictions (Boyd et al., 1978).
Indwelling electrodes are influenced not only by waves that actually pass by their conducting
surfaces but also by waves that pass within a few millimeters of the bare conductor. The same is
true for surface electrodes. The field equations that describe the electrode potential were origi-
nally derived by Lorente de No (1947) and were extended in rigorous formulations by Plonsey
(1964, 1974) and Rosenfalck (1969). These equations were complicated by the formulation of
current density functions to describe the temporal and spatial depolarization and repolarization
of the muscle membrane. Thus, simplification of the current density to a dipole or tripole
(Rosenfalck, 1969) yielded a reasonable approximation when the active fiber was more than
1 mm from the electrode surface (Andreassen and Rosenfalck, 1981).
In the dipole model (see Figure 9.2), the current is assumed to be concentrated at two points
along the fiber: a source of current, I, representing the depolarization, and a sink of current, −I,
representing the repolarization, both separated by a distance b. The potential Φ, recorded at a
point electrode at a distance r from the current source, is given by:
I 1
Φ= (9.1)
4IIσ r
where σ is the conductivity of the medium, which is assumed to be isotropic (uniform in all spatial
directions).
The net potential recorded at the point electrode from both source and sink currents is:
I 1 I 1
Φ= −
4IIσ r 1 4IIσ r 2
(9.2)
I 1 1
= −
4IIσ r 1 r 2
where r1 and r2 are the distances to the source and sink currents.
Electrode
r2 r1
I I Muscle fiber
t1 t2 t3 t4 t5
Figure 9.2 Propagation of motor unit action potential wave front as it passes beneath a recording electrode on the
skin surface. The electrode voltage is a function of the magnitude of the
electrode to the depolarizing and repolarizing currents.
ISTUDY
The time history of the action potential then depends on r1 and r2, which vary with time as the
wave propagates along the muscle fiber. At t1, as the wave approaches the electrode (r1 < r2), the
potential will thus be positive and will increase. It will reach a maximum at t2; then as r1 becomes
nearly equal to r2, the amplitude suddenly decreases and passes through zero at t3 when the dipole
is directly beneath the electrode (r1 = r2). Then it becomes negative as the dipole propagates away
from the electrode (r1 > r2). Thus, a biphasic wave is recorded by a single electrode.
A number of recording and biological factors affect the magnitude and shape of the biphasic
signal. The duration of each phase is a function of the propagation velocity, the distance b (which
varies from 0.5 to 2.0 mm) between source and sink, the depth of the fiber, and the electrode
surface area. Equation (10.2) assumes a point electrode. Typical surface electrodes are not point
electrodes but have a finite surface area, and each point on the surface can be considered an area
of point sources; the potential on the surface is the average of all point-source potentials.
Figure 9.3 is presented from a study by Fuglevand et al. (1992) to demonstrate the influence
of the electrode size and shape on the electrode potential. Two different orientations of strip elec-
trodes are shown: (a) along an axis parallel to the muscle fiber – and (d) at right angles to the
muscle fiber. (b) and (e) show the location of a strip of point-source electrodes to represent these
two strip electrodes, each 10 mm long with 10 point sources; (c) and (f) show the action potentials
at each point source. For the strip electrode in (a), the distance from the points on the electrode to
the fiber, rf, is constant; thus, each action potential has the same amplitude and shape. However,
there is a phase difference between each of the point-source action potentials such that the aver-
age action potential is lower and longer than the individual action potentials. For the strip elec-
trode in (d), the distance rf is different for every point source; thus, each action potential decreases
as rf increases, as is seen in (f). The phase of each point-source action potential has the same zero-
crossing time, but the smaller action potentials from the further point sources are slightly longer in
duration. Finally, (g) compares the net action potentials from a point source, strip (a), strip (d),
and a square electrode 10 × 10 mm2. As is evident from these action potentials, it is desirable to
use electrodes with as small a surface area as possible. Also, as will be seen in Section 9.2.5, the
smaller surface area electrodes are less influenced by cross-talk.
The zone of pick-up for individual fibers by an electrode depends on two physiological vari-
ables, I and r. Larger muscle fibers have larger dipole currents, which increase with the diameter
of the fiber. For an m.u.a.p., the number of fibers innervated must also be considered. The poten-
tial detected from the activation of a motor unit will be equivalent to the sum of the constituent
fiber potentials. Therefore, the pick-up zone for a motor unit that innervates a large number of
fibers will be greater than that for a unit with fewer fibers. An estimate of the detectable pick-up
distance for small motor units (50 fibers) is about 0.5 cm and, for the largest motor units (2500
fibers), about 1.5 cm (Fuglevand et al., 1992). Thus, surface electrodes are limited to detecting
m.u.a.p.’s from those fibers quite close to the electrode site and are not prone to pick up from
adjacent muscles (the phenomenon known as cross-talk) unless the muscle being recorded from
is very small.
Most EMGs require two electrodes over the muscle site, so the voltage waveform that is
recorded is the difference in potential between the two electrodes. Figure 9.4 shows that the volt-
age waveform at each electrode is almost the same but is shifted slightly in time. Thus, a biphasic
potential waveform at one electrode usually results in a different signal that has three phases. The
closer the spacing between the two electrodes, the more the difference signal looks like a time
differentiation of that signal recorded at a single electrode (Kadefors, 1973). Thus, closely spaced
electrodes result in the spectrum of the EMG having higher-frequency components than those
recorded from widely spaced electrodes.
9.1.4 Duration of the Motor Unit Action Potential

As indicated in the previous section, the larger the surface area, the longer the duration of the
m.u.a.p. Thus, surface electrodes automatically record
ISTUDY
9.1 ELECTROPHYSIOLOGY OF MUSCLE CONTRACTION 193
(a) (d)
Axon Axon
Electrode
X Electrode X
rf1
–I –I
I I
rf rf10
Z Z
Y Y
(b) (e)
Z Z
Axon Axon r2
r2
rf rf
r1 r1
–I I –I I
Muscle fiber Muscle fiber
(c) (f)
1 μV 5 μV
0.01 s
0.01 s
(g)
Point electrode
z-distributed electrode
x-distributed electrode
Square electrode
5 μV
0.01 s
Figure 9.3 Influence of electrode geometry on the predicted action potentials from a single electrode with a variety
of shapes and sizes. The potential was computed as the algebraic mean of the potentials detected by an array of point-
source electrodes. See the text for detailed discussion. (Reproduced by permission from Biological Cybernetics,
Fuglevand et al. (1992). With kind permission of Springer Science + Business Media.)
electrodes (Kadefors, 1973; Basmajian, 1973). Needle electrodes record durations of 3–20 ms,
while surface electrodes record durations about twice that long. However, for a given set of elec-
trodes, the duration of the m.u.a.p.’s is a function of the velocity of the propagating wave front
and the depth of the motor unit below the electrode’s surface. The velocity of propagation of the
m.u.a.p. in normals has been found to be about 4 m/s
ISTUDY
Va
Vb
Va–Vb
Figure 9.4 Voltage waveform present at two electrodes because of a single propagating wave. Voltage recorded is
the difference voltage Va − Vb, which is triphasic in comparison with the biphasic waveform seen at a single
electrode.
velocity, the shorter the duration of the m.u.a.p. Such a relationship has been used to a limited
extent in the detection of velocity changes. In fatigue and in certain myopathies (muscle pathol-
ogies), the average velocity of the m.u.a.p.’s that are recruited is reduced; thus, the duration of the
m.u.a.p.’s increases (Johansson et al., 1970; Gersten et al., 1965; Kadefors, 1973). Because the
peak amplitude of each phase of the m.u.a.p. remains the same, the area under each phase will
increase. Thus, when we measure the average amplitude of the EMG (from the full-wave rectified
waveform), it will appear to increase (Fuglevand, et al., 1992). During voluntary contractions, it
has been possible, under special laboratory conditions (Milner-Brown and Stein, 1975), to detect
the duration and amplitude of muscle action potentials directly from the EMG. However, during
unconstrained movements, a computer analysis of the total EMG would be necessary to detect a
shift in the frequency spectrum (Kwatny et al., 1970), or an autocorrelation analysis can yield the
average m.u.a.p. duration (Person and Mishin, 1964).
The distance between the motor unit and the surface of the electrode severely influences the
amplitude of the action potential, as is predicted in Equation (10.2), and also influences the dura-
tion of the action potential. Figure 9.5, taken from Fuglevand et al. (1992), predicts the duration
and amplitude of the action potential from a 50-fiber motor unit with electrode-unit distances out
to 20 mm. It is evident that the duration of the action potential increases as the amplitude
decreases with distance. Thus, the frequency content of the m.u.a.p.’s from the more distant sites
decreases. Fuglevand et al. (1992) showed that the mean power frequency (MPF) of the m.u.a.p.
would decrease from 160 Hz at an electrode-motor unit distance of 1 mm to 25 Hz at an electrode-
motor unit distance of 20 mm.
9.1.5 Detection of Motor Unit Action Potentials from Electromyogram During Graded
Contractions
Using recordings from multiple indwelling electrodes, DeLuca et al. (1982) pioneered techniques
to identify individual m.u.a.p.’s during a low-level graded contraction. More recent techniques
use a quadrifilar needle electrode recording of three channels followed by decomposition algo-
rithms involving template matching, template updating, and motor unit firing statistics (DeLuca,
1993). As many as four motor units were detected and their firing rates were tracked from 0 to
100% of maximum voluntary contraction (MVC) of the tibialis anterior (TA) muscle (Erim et al.,
1996). It should be noted that only a fraction of all active
ISTUDY
9.2 RECORDING OF THE ELECTROMYOGRAM 195
400 μV
200
10 ms
Mean power frequency (Hz) 150
10 μV
100
10 ms 1 μV
10 ms
50
0
0 5 10 15 20
Electrode–motor unit territory distance (mm)
Figure 9.5 Variation in the amplitude and frequency content of the motor unit action potential with increased
electrode-motor unit distance. Predictions were for a small motor unit (50 fibers) with 4-mm2 bipolar electrodes
with an interelectrode spacing of 11 mm. Shown are the m.u.a.p.’s at 1-, 10-, and 20-mm distances.
(Reproduced by permission from Biological Cybernetics, Fuglevand et al. (1992). With kind permission of
Springer Science + Business Media.)
within the pick-up region of the electrode array. EMG recordings from surface electrodes have
been decomposed in order to determine the firing profile of those motor units detected (McGill
et al., 1987).
9.2 RECORDING OF THE ELECTROMYOGRAM
A biological amplifier of certain specifications is required for the recording of the EMG, whether
from surface or from indwelling electrodes. It is valuable to discuss the reasons behind these
specifications with respect to considerable problems in getting a “clean” EMG signal. Such a
signal is the summation of m.u.a.ps and should be undistorted and free of noise or artifacts.
Undistorted means that the signal has been amplified linearly over the range of the amplifier
and recording system. The larger signals (up to 5 mV) have been amplified as much as the smaller
signals (100 μV and below). The most common distortion is overdriving of the amplifier system
such that the larger signals appear to be clipped off. Every amplifier has a dynamic range and
should be such that the largest EMG signal expected will not exceed that range. Noise can be
introduced from sources other than the muscle and can be biological in origin or man-made.
An electrocardiogram (ECG) signal picked up by EMG electrodes on the thoracic muscles
can be considered unwanted biological noise. Man-made noise usually comes from power lines
(60 Hz in the United States or 50 Hz in Europe) or from electronics or is generated within the
components of the amplifier. Motion artifacts generally refer to false signals generated by the
electrodes themselves or the cabling system. Therefore, care should be taken to secure wires
and electrodes to the participant so that movement near the electrode detection sites would
not be affected (Figure 9.6).
The major considerations to be made when specifying the EMG amplifier are:
1. Gain and dynamic range

2. Input impedance
ISTUDY
Figure 9.6 Surface EMG electrodes secured with adhesive material to a patient to prevent movement artifact of the
wires and electrode.
3. Frequency response
4. Common-mode rejection
9.2.1 Amplifier Gain

Surface EMGs have a maximum amplitude of 5 mV peak to peak, as recorded during a MVC.
Indwelling electrodes can have a larger amplitude, up to 10 mV. A single m.u.a.p. has an ampli-
tude of about 100 μV. The noise level of the amplifier is the amplitude of the higher-frequency
random signal seen when the electrodes are shorted together, and should not exceed 50 μV, pref-
erably 20 μV. The gain of the amplifier is defined as the ratio of the output voltage to the input
voltage. For a 2-mV input and a gain of 1000, the output will be 2 V. The exact gain chosen for
any given situation will depend on what is to be done with the output signal. EMG had tradition-
ally been viewed and recorded on a pen recorder, but nearly all modern units will transfer the data
into digital format and be stored on a computer. The amplified EMG should not exceed the range
of input signals expected by the recording equipment. Fortunately, most of these recording sys-
tems have internal amplifiers that can be adjusted to accommodate a wide range of input signals.
In general, a good bioamplifier should have a range of gains selectable from 100 to 10,000. Inde-
pendent of the amplifier gain, the amplitude of the signal should be reported as it appears at the
electrodes, in millivolts.
9.2.2 Input Impedance

The input impedance (resistance) of a biological amplifier must be sufficiently high so as not to
attenuate the EMG signal as it is connected to the input terminals of the amplifier. Consider the
amplifier represented in Figure 9.7a. The active input terminals are 1 and 2, with a common ter-
minal c. The need for a three-input terminal amplifier (differential amplifier) is explained in
Section 9.2.4.
Each electrode-skin interface has a finite impedance that depends on many factors: thickness
of the skin layer, cleaning of the skin prior to the attachment of the electrodes, area of the elec-
trode surface, and temperature of the electrode paste (it warms up from room temperature after
ISTUDY
(a)
Current
from
muscles 1
To A/D
Biological
Ri amplifier Vout converter and
computer
Current to
muscles 2
Common
Silver
Electrode chloride disc
paste Insulating
disc
(b)
Rs1 i
1
Vemg Vin Ri Biological Vout

amplifier
2
Internal EMG signal i
Rs2
(just under surface
of skin)
Figure 9.7 Biological amplifier for recording electrode potentials. (a) Current resulting from muscle action
potentials flows across skin – electrode interface to develop a voltage Vin at the input terminals of the amplifier.
A third, common electrode is normally required because the amplifier is a differential amplifier. (b) Equivalent
circuit showing electrodes replaced by series resistors Rs1 and Rs2.V in will be nearly equal to VEMG if Ri >> Rs.
attachment). Indwelling electrodes have a higher impedance because of the small surface area of
bare wire that is in contact with the muscle tissue.
In Figure 9.7b, the electrode-skin interface has been replaced with an equivalent resistance.
This is a simplification of the actual situation. A correct representation is a more complex imped-
ance to include the capacitance effect between the electrode and the skin. As soon as the amplifier
is connected to the electrodes, the minute EMG signal will cause current to flow through the
electrode resistances Rs1 and Rs2 to the input impedance of the amplifier Ri. The current flow
through the electrode resistance will cause a voltage drop so that the voltage at the input terminals
Vin will be less than the desired signal VEMG. For example, if Rs1 = Rs2 = 10,000 Ω and
Ri = 80,000 Ω, a 2-mV EMG signal will be reduced to 1.6 mV at Vin. A voltage loss of
0.2 mV occurs across each of the electrodes. If Rs1 and Rs2 were decreased by better skin
preparation to 1000 Ω, and Ri was increased to 1 M Ω, the 2-mV EMG signal would be reduced
only slightly, to 1.998 mV. Thus, it is desirable to have input impedances of 1 M Ω or higher and
to prepare the skin to reduce the impedance to 1000 Ω or less. For indwelling electrodes, the
electrode impedance can be as high as 50, 000 Ω, so an amplifier with at least 5 M Ω input imped-
ance should be used.
9.2.3 Frequency Response

The frequency bandwidth of an EMG amplifier should be such as to amplify, without attenuation,
all frequencies present in the EMG. The bandwidth of any amplifier, as shown in Figure 9.8, is the
difference between the upper cutoff frequency f2 and the lower cutoff frequency f1. The gain of
ISTUDY
1000 60
Gain
800 (dB)
Gain
57
600
400 54
Bandwidth
f1 f2
Frequency, Hz
Figure 9.8 Frequency response of the biological amplifier showing a gain of 1000 (60 dB), and lower and upper
cutoff frequencies f1 and f2.
the amplifier at these cutoff frequencies is 0.707 of the gain in the midfrequency region. If we
express the midfrequency gain as 100%, the gain at the cutoff frequencies has dropped to 70.7%,
or the power has dropped to (0.707)2 = 0.5. These are also referred to as the half-power points.
Often amplifier gain is expressed in logarithmic form and expressed in decibels,
gain dB = 20 log10 linear gain (9.3)
If the linear gain were 1000, the gain in decibels would be 60, and the gain at the cutoff fre-
quency would be 57 dB (3 dB less than that at midfrequency).
In a high-fidelity amplifier used for music reproduction, f1 and f2 are designed to accommodate
the range of human hearing, from 50 to 20,000 Hz. All frequencies present in the music will be
amplified equally, producing a true undistorted sound at the loudspeakers. Similarly, the EMG
should have all its frequencies amplified equally. The spectrum of the EMG has been widely
reported in the literature, with a range from 5 Hz at the low end to 2000 Hz at the upper end.
For surface electrodes, the m.u.a.p.’s are longer in duration and, thus, have negligible power
beyond 1000 Hz. A recommended range for surface EMG is 10–1000 Hz and 20–2000 Hz for
indwelling electrodes. If computer pattern recognition of individual m.u.a.p.’s is being done,
the upper cutoff frequencies should be increased to 5 and 10 kHz, respectively. Figure 9.9 shows
a typical EMG spectrum, and it can be seen that most of the signal is concentrated in the band
between 20 and 200 Hz, with a lesser component extending up to 1000 Hz.
The spectra of other physiological and noise signals must also be considered. ECG signals
contain power out to 100 Hz, so it may not be possible to eliminate such interference, especially
EMG spectral density

Power/Frequency
indwelling
surface
0 200 400 600 800 1000

Frequency
Figure 9.9 Frequency spectrum of EMG as recorded via surface and indwelling electrodes. The higher-frequency
content of indwelling electrodes is the result of the closer spacing between electrodes and their closer proximity to
the active muscle fibers.
ISTUDY
(a) Raw EMG (b)

Raw EMG
10–1000 Hz 500 ms 10–1000 Hz 500 ms
30–1000
10–30
50–1000
10–50
100–1000
10–100
200–1000
10–200
300–1000
10–400
Raw EMG
400–1000
Figure 9.10 Surface EMG filtered with varying cutoff frequencies. (a) With lower cutoff frequency varied from
30 to 400 Hz, showing the effect of rejecting lower frequencies. (b) With upper cutoff frequency varied from 30 to
400 Hz, showing the loss of higher frequencies.
when monitoring muscle activity around the thoracic region. The major interference comes from
power line hum: in North America it is 60 Hz, in Europe 50 Hz. A notch filter can be applied to
remove that source of noise, however physiologic EMG at the frequency will also be removed.
Movement artifacts, fortunately, lie in the 0–10 Hz range and normally should not cause
problems. Unfortunately, some of the lower-quality cabling systems can generate large
low-frequency artifacts that can seriously interfere with the baseline of the EMG recording.
Usually, such artifacts can be eliminated by good low-frequency filtering, by setting f1 to about
20 Hz. If this fails, the only solution is to replace the cabling or use an EMG system that includes
microamplifiers at the skin surface.
It is valuable to see the EMG signal as it is filtered using a wide range of bandwidths.
Figure 9.10 shows the results of such filtering and the obvious distortion of the signal when
f1 and f2 are not set properly.
9.2.4 Common-Mode Rejection

The human body is a good conductor and, therefore, will act as an antenna to pick up any elec-
tromagnetic radiation that is present. The most common radiation comes from domestic power:
power cords, fluorescent lighting, and electric machinery. The resulting interference may be so
large as to prevent recording of the EMG. If we were to use an amplifier with a single-ended
input, we would see the magnitude of this interference. Figure 9.11 depicts hum interference
on the active electrode. It appears as a sinusoidal signal, and if the muscle is contracting, its
EMG is added. However, hum could be 100 mV and would drown out even the largest EMG
signal.
If we replace the single-ended amplifier with a differential amplifier (see Figure 9.12), we can
possibly eliminate most of the hum. Such an amplifier takes the difference between the signals on
the active terminals. As can be seen, this hum interference appears as an equal amplitude on both
active terminals. Because the body acts as an antenna, all
ISTUDY
EMG + hum
1
EMG + hum
Amplifier
VOUT
No hum on ground
connection
Figure 9.11 Single-ended amplifier showing lack of rejection of hum present on ungrounded active terminal.
Vhum + emg1
e1
Gain = A
VOUT = A (e1–e2)
e2
Vhum + emg2
Figure 9.12 Biological amplifier showing how the differential amplifier rejects the common (hum) signal by
subtracting the hum signal that is present at an equal amplitude at each active terminal. Different EMG signals
are present at each electrode; thus, the subtraction does not result in a cancellation.
Because this unwanted signal is common to both active terminals, it is called a common-mode
signal. At terminal 1, the net signal is V hum + emg1; at terminal 2 it is V hum + emg2. The amplifier
has a gain of A. Therefore, the ideal output signal is:
eo = A e 1 − e2
= A V hum + emg1 − V hum − emg2 (9.4)
= A emg1 − emg2
The output eo is an amplified version of the difference between the EMGs on electrodes 1
and 2. No matter how much hum is present at the individual electrodes, it has been removed
by a perfect subtraction within the differential amplifier. Unfortunately, a perfect subtraction
never occurs, and the measure of how successfully this has been done is given by the
ISTUDY
common-mode rejection ratio (CMRR). If CMRR is 1000 : 1, then all but 1/1000 of the hum will
be rejected. Thus, the hum at the output is given by:
A × V hum
V o hum = (9.5)
CMRR
Example 9.1. Consider the EMG on the skin to be 2 mV in the presence of hum of 500 mV.
CMRR is 10,000:1 and the gain is 2000. Calculate the output EMG and hum.
eo = A emg1 − emg2
= 2000 × 2 MV = 4 V
output hum = 2000 × 500 mV 10, 000 = 100 mV
Hum is always present to some extent unless the EMG is being recorded with battery-powered
equipment not in the presence of domestic power. Its magnitude can be seen in the baseline when
no EMG is present. Figure 9.13 shows two EMG records, the first with hum quite evident, the
latter with negligible hum.
(a) With hum present.

(b) Without hum.
CMRR is often expressed as a logarithmic ratio rather than a linear ratio. The units of this ratio
are decibels,
CMRR dB = 20log10 CMRR linear (9.6)
(a)
500 ms
(b)
Figure 9.13 Recordings of an EMG signal.
ISTUDY
If CMRR = 10, 000 : 1, then CMRR (dB) = 20log1010, 000 = 80 dB. In good-quality biolog-
ical amplifiers, CMRR should be 80 dB or higher.
9.2.5 Cross-Talk in Surface Electromyograms

The detectable pick-up distance was estimated to be about 0.5 cm for small motor units and about
1.5 cm for the largest units (Fuglevand et al., 1992). In Figure 9.14, for example, we see a number
of surface electrodes over several thigh muscles in the midthigh region. The range of pick-up is
shown as an arc beneath each electrode. Those m.u.a.p.’s whose fibers are close to each electrode
are not subject to cross-talk; however, there is an overlapping zone of pick-up where both elec-
trodes detect m.u.a.p.’s from the same active motor units. This common pick-up is called
cross-talk.
Recommendations for the placement of surface electrodes have been established and pub-
lished through the project “Surface ElectroMyoGraphy for the Non-Invasive Assessment of Mus-
cles” (SENIAM). These sites were chosen to maximize signal amplitude from the desired muscle
while reducing cross-talk from adjacent muscles. However, because of close proximity of the
desired muscles in many research applications, it may be necessary to check that cross-talk is
minimal. The most common way to test for cross-talk is to conduct a manual resistance test prior
to any given experiment. Consider two adjacent muscles A and B that, during the course of a
given movement, have common periods of activity. The purpose of the manual resistance tests
is to get a contraction on muscle A without any activity on B and vice versa. Consider electrodes
placed on the leg on three muscles that are fairly close together: TA, peroneus longus (PER), and
Rectus femoris
Vastus medialis Vastus lateralis
Indwelling
Sartorius
Vastus intermedius
Gracilis
Biceps femoris
Semimembranosus Adductor magnus
Semitendinosus
Figure 9.14 Location of seven surface electrodes placed across the midthigh showing the region of pick-up for
each site relative to the underlying muscles. The zone of pick-up for each is shown as an arc beneath electrode; there
is an overlapping zone between adjacent electrodes that will result in some cross-talk. Deeper muscles, such as the
vastus intermedius and adductor magnus, require indwelling electrodes whose zone of pick-up is small because of
the small surface area and close spacing of the electrode tips.
ISTUDY
lateral gastrocnemius (LG). Functionally, the TA is a dorsiflexor and invertor, PER is an evertor
and plantarflexor, and LG is a plantarflexor. The results of a functional test for cross-talk from
the PER to the TA and LG is presented in Figure 9.15. Because it is difficult for many subjects to
voluntarily create a unique contraction (such as eversion for the PER), it is very helpful for
the subject to view the EMG recording and use visual feedback to assist in the test. Here, we
see two very distinct PER contractions. The first was eversion with a small amount of dorsiflex-
ion: note the minor activity on TA and negligible amount on LG. The second contraction was
eversion with a minor amount of plantarflexion: note the minor activity on the LG and none
on TA. It is not critical to be successful on both tests simultaneously, or even on the very first
attempt. The student is referred to Winter et al. (1994) for more details and examples of manual
resistance tests.
However, there are many situations when a manual resistance test cannot be done, and there is
a good chance that adjacent channels may have some common EMG signal. The question is how
much do they have in common? This question was initially addressed by Winter et al. (1994)
using the well-recognized signal-processing technique called cross-correlation (see
Section 2.1). The adjacent channels, x(t) and y(t), are isometrically contracted for about 10
TA
PER
LG
Figure 9.15 Results of a manual resistance test to ensure no cross-talk between the peroneii (PER) muscle and the
tibialis anterior (TA) and lateral gastrocnemius (LG) muscles. The first trial showed negligible cross-talk between
PER and TA; the second contraction showed negligible cross-talk between PER and LG. See text for further details.
(Reprinted from the Journal of Electromyography and Kinesiology, Winter et al. (1994), with permission of
Elsevier.)
ISTUDY
seconds at a moderate contraction level. The general equation for cross-correlation, Rxy(τ),
between signals x(t) and y(t) is:
T
1
T xt y t − τ dt
Rxy τ = 0
(9.7)
Rxx 0 Ryy 0
where x(t) and y(t), respectively, are at a phase shift of 0. Rxx(0) and Ryy(0) are the mean squares of
x(t) and y(t), respectively, over the time T . τ is normally varied ±30 ms, and a typical cross-
correlation for two electrode pairs located 2 cm apart is presented in Figure 9.16. Note that
the peak correlation is 0.6 at zero phase shift. Because the electrode pairs were placed parallel
with the muscle fibers, each pair was the same distance from the motor point; thus, any common
m.u.a.p.’s will be virtually in phase. With the peak Rxy = 0.6, the net correlation, R2xy = 0 36, tells
us the two EMG had 36% of their signals in common. Rxy drops off sharply as the distance
between electrode pairs is increased (Winter et al., 1994). At 2.5 cm, Rxy = 0.48 (23% cross-talk);
at 5.0 cm, Rxy = 0.24 (6% cross-talk); and at 7.5 cm, Rxy = 0.14 (2% cross-talk). Reducing the size
of the electrodes reduces the pick-up zone of the electrode, and thus the cross-talk. The
motor units at a greater distance from the electrodes contribute most to the cross-talk, and
the m.u.a.p.’s from those units are longer duration (lower frequency). Thus, by processing the
EMG through a differentiator, the higher-frequency (closer) m.u.a.p.’s are emphasized, while
the lower-frequency (further) m.u.a.p.’s are attenuated.
The double differentiation technique can be used to either remove or identify if cross-talk is
occurring during an EMG recording (DeLuca, 1997). The use of an electrode with 3 detection
surfaces is required to perform this technique which differentiations the EMG signal between
the 1 and 2 detection surfaces and the 2 and 3 detection surfaces to cancel out very distant signals
likely coming from further away muscles. Although useful this approach is not commonly used in
practice. Students should spend time perfecting electrode placement within the mid muscle belly
to avoid cross-talk and accurate and reliable EMG signals. Ultrasound can be used during training
to help with muscle identification and electrode placement.
0.6
0.5
0.4
0.3
0.2
Rxy
0.1
–0.1
–0.2
–0.3
–0.4
–30 –20 –10 0 10 20 30
Time shift (ms)
Figure 9.16 Results of a cross-correlation, Rxy, between two adjacent electrodes in Figure 9.13. Here, Rxy = 0.6,
indicating a common pick-up of R2xy = 0 36, or 36% cross-talk. (Reprinted from the Journal of Electromyography
and Kinesiology, Winter et al. (1994), with permission of Elsevier.)
ISTUDY
9.3 PROCESSING OF THE ELECTROMYOGRAM 205
9.2.6 Recommendations for Surface Electromyogram Reporting

and Electrode Placement Procedures
Over the past 30 years, the availability of EMG recording equipment and the number of labora-
tories using such equipment has exploded dramatically. Most labs have developed their own pro-
tocols regarding the details needed for reporting results and for the selection of the electrode sites
over the muscles. The first attempt at developing recommended standards was done by the Inter-
national Society of Electrophysiological Kinesiology, which resulted in an Ad Hoc Committee
Report (Winter et al., 1980). More recently, a considerably more in-depth report was published
with the support of the European Union and was called SENIAM (Surface EMG for a Non-
Invasive Assessment of Muscles). During 1996–1999, the report was developed and reviewed
by over 100 laboratories (Hermens et al., 2000), and the final report was published as a booklet
and as a CD-ROM: SENIAM 8: European Recommendations for Surface Electromyogra-
phy, 1999.
The SENIAM recommendations as to information to be reported are as follows:
1. Electrode size, shape (square, circular, etc.), and material (Ag, AgCl, Ag/AgCl, etc.)
2. Electrode type (monopolar, bipolar, one- or two-dimensional array)
3. Skin preparation, interelectrode distance
4. Position and orientation on the muscle (recommendations for 27 muscles are included)
The location of electrode placement can have a dramatic effect of the recorded EMG sig-
nal. Electrodes should be placed at the mid muscle belly (transverse and sagittal plane).
When a muscle contracts the distal and proximal attachment will be brought to the middle
of the muscle.
Therefore, by placing the electrodes at the mid muscle belly there will be minimal movement
of the motor units being recorded under the electrodes. Recording a similar group of motor
units throughout a range of motion will provide a valid amplitude and frequency profile, which
can be used with confidence to identify neuromuscular characteristics of a given muscle. If an
electrode is placed closer to either tendon end there will be a smaller pool of motor units repre-
sented in the EMG signal and that pool will change during the contraction of the muscle leading
to inaccurate conclusions about the neuromuscular behavior of the muscle. If the electrode is
placed medial or lateral to the sagittal plane of the muscle there is potential for cross-talk to
occur as described previously. Another consideration for electrode placement is that the detec-
tion surfaces are oriented perpendicular to the direction of muscle fibers. If the detection sur-
faces are off angle to the muscle fibers the conduction velocity will be affected and the
amplitude and frequency of the signal profile will be altered. For these reasons, it is important
to have a strong knowledge of muscular anatomy prior to performing electrode placements. In
addition, palpation techniques of surface anatomy should be practiced to enhance electrode
placement accuracy.
9.3 PROCESSING OF THE ELECTROMYOGRAM
Once the EMG signal has been amplified, it can be processed for comparison or correlation with
other physiological or biomechanical signals. Given the high frequency nature of the raw EMG
and typical mean value of zero, additional processing makes it better suited for quantitative clin-
ical and research applications.
The more common types of online processing are:
1. Half- or full-wave rectification (the latter also called absolute value)

2. Linear envelope detector (half- or full-wave rectifier
ISTUDY
3 mv
Raw EMG
BIOAMP.
1
3 mv
Full wave
2
rectify
1 mv
L. P. Filer Linear
3
envelope
1 mv. s
4
Integrate
over contr.
50 μv. s
5
Integrate 50 ms
then reset
6
Integrate to
present voltage
then reset Each pulse = 38 μv.S
7
Figure 9.17 Schematic diagram of several common EMG processing systems and the results of simultaneous
processing of EMG through these systems. See the text for details.
3. Integration of the full-wave rectified signal over the entire period of muscle contraction
4. Integration of the full-wave rectified signal for a fixed time, reset to zero, then integration
cycle repeated
5. Integration of the full-wave rectified signal to a preset level, reset to zero, then integration
repeated
These various processing methods are shown schematically in Figure 9.17, together with a
sample record of a typical EMG processed in all five ways. This is now discussed in detail.
9.3.1 Full-Wave Rectification

The full-wave rectifier generates the absolute value of the EMG, usually with a positive polarity.
The original raw EMG has a mean value of zero because it is recorded with a biological amplifier
with a low-frequency cutoff around 10 Hz. However, the full-wave rectified signal does not cross
through zero and, therefore, has an average or bias level that fluctuates with the strength of the
muscle contraction. The quantitative use of the full-wave rectified signal by itself is somewhat
limited; it serves as an input to the other processing schemes. The main application of the full-
wave rectified signal is in semi-quantitative assessments of the phasic activity of various muscle
groups. A visual examination of the amplitude changes of the full-wave rectified signal gives a
good indication of the changing contraction level of the muscle. The proper unit for the amplitude
of the rectified signal is the millivolt, the same as for the original EMG.
ISTUDY
9.3 PROCESSING OF THE ELECTROMYOGRAM 207
9.3.2 Linear Envelope

If we filter the full-wave rectified signal with a low-pass filter, we have what is called the linear
envelope. It can be described best as a moving average because it follows the trend of the EMG
and quite closely resembles the shape of the tension curve. It is reported in millivolts. There is
considerable confusion concerning the proper name for this signal. Many researchers call it an
integrated EMG (IEMG). Such a term is quite wrong because it can be confused with the math-
ematical term integrated, which is a different form of processing.
There is a need to process the signal to provide the assessor with a pattern that can be justified
on some biophysical basis. Some researchers have full-wave rectified the raw EMG and low-pass
filtered at a high frequency but with no physiological basis (Forssberg, 1985; Murray et al.,
1985). If it is desired that the linear envelope bear some relationship to the muscle force or
the joint moment of force, the processing should model the biomechanics of muscle tension gen-
eration. The basic unit of muscle tension is the muscle twitch, and the summation of muscle
twitches as a result of recruitment is matched by a superposition of m.u.a.p.’s. There is an inherent
delay between the m.u.a.p. and the resultant twitch waveform. If we consider the full-wave rec-
tified signal to be an impulse and the twitch to be the response, we can define the transfer function
of the desired processing. The duration of the full-wave rectified m.u.a.p. is about 10 ms, while
the twitch waveform peaks at 50–110 ms and lasts up to 300 ms, so this impulse/response
relationship is close. Twitch waveforms have been analyzed and have been found to be a
second-order system, critically damped or slightly overdamped (Crosby, 1978; Milner-Brown
et al., 1973). The cutoff frequency of these second-order responses ranged from 2.3 to 7.8 Hz.
Figure 9.18 is presented to show the linear envelope processing of the EMG to model the
relationship between the m.u.a.p. and the twitch. A critically damped second-order low-pass filter
has a cutoff frequency fc, which is related to the twitch time T as follows:
f c = 1 2πT (9.8)
Thus, the table in Figure 9.18 shows the relationship between fc and T for the range of twitch
times reported in the literature. The soleus muscle with a twitch time ≈106 ms would require a
Linear envelope
EMG processing
Bio-amp
Low-pass filter (critically damped)

Frequency response Impulse fc (Hz) T (ms)
1.0 T
response 4 40
0.707 1
T= 3 53
2π fc
2 80
fc 1.5 106
Figure 9.18 Linear envelope processing of the EMG with a critically damped low-pass filter to match the impulse
response of the muscle being recorded. The full-wave rectified EMG acts as a series of impulses that, when filtered,
mimic the twitch response of the muscle and, in a graded contraction, mimic the superposition of muscle twitches.
ISTUDY
filter with fc = 1.5 Hz. The reader is referred back to Section 8.0.5 for information on the twitch
shape and times.
Good correlations have been reported between the muscle force and linear envelope wave-
forms during isometric anisotonic contractions (Calvert and Chapman, 1977; Crosby, 1978).
Muscle contraction can also be modeled as a mass-spring-damper system that was critically
damped. The critically damped low-pass filters described earlier have exactly the same response
as this previously mechanical damper system. Thus, the matching of the muscle force waveform
with that from the model is exactly the same as would be achieved with a linear envelope detector
with the same cutoff frequency as used in the mechanical model.
9.3.3 True Mathematical Integrators

There are several different forms of mathematical integrators, as shown in Figure 9.17. The pur-
pose of an integrator is to measure the “area under the curve.” Thus, the integration of the full-
wave rectified signal will always increase as long as any EMG activity is present. The simplest
form starts its integration at some preset time and continues during the time of the muscle activity.
At the desired time, which could be a single contraction or a series of contractions, the integrated
value can be recorded. The unit of a properly integrated signal is millivolt-seconds (mV s). The
only true way to find the average EMG during a given contraction is to divide the integrated value
by the time of the contraction; this will yield a value in millivolts.
A second form of integrator involves a resetting of the integrated signal to zero at regular inter-
vals of time (40–200 ms). Such a scheme yields a series of peaks that represent the trend of the
EMG amplitude with time. Each peak represents the average EMG over the previous time inter-
val, and the series of peaks is a “moving” average. Each peak has units of millivolt-seconds so
that the sum of all the peaks during a given contraction yields the total integrated signal, as
described in the previous paragraph. There is a close similarity between this reset, integrated
curve, and the linear envelope. Both follow the trend of muscle activity. If the reset time is
too high, it will not be able to follow rapid fluctuations of EMG activity, and if it is reset too
frequently, noise will be present in the trendline. If the integrated peaks are divided by the inte-
gration time, the amplitude of the signal can again be reported in millivolts.
A third common form of integration uses a voltage level reset. The integration begins before
the contraction. If the muscle activity is high, the integrator will rapidly charge up to the reset
level; if the activity is low, it will take longer to reach rest. Thus, the strength of the muscle con-
traction is measured by the frequency of the resets. A high frequency of reset pulses reflects high
muscle activity; a low frequency represents low muscle activity. Intuitively, such a relationship is
attractive to neurophysiologists because it has a similarity to the action potential rate present in
the neural system. The total number of counts over a given time is proportional to the EMG activ-
ity. Thus, if the threshold voltage level and the gain of the integrater are known, the total EMG
activity (millivolt-seconds) can be determined.
9.4 RELATIONSHIP BETWEEN ELECTROMYOGRAM

AND BIOMECHANICAL VARIABLES
The major reason for processing the basic EMG is to derive a relationship between it and some
measure of muscle function. A question that has been posed for years is: “How valuable is the
EMG in predicting muscle tension?” Such a relationship is very attractive because it would give
an inexpensive and noninvasive way of monitoring muscle tension. Also, there may be informa-
tion in the EMG concerning muscle metabolism, power, fatigue state, or contractile elements
recruited.
ISTUDY
9.4 RELATIONSHIP BETWEEN ELECTROMYOGRAM AND BIOMECHANICAL VARIABLES 209
9.4.1 Electromyogram Versus Isometric Tension

Bouisset (1973) has a seminal review of the state of knowledge regarding the EMG and muscle
tension in normal isometric contractions. The EMG processed through a linear envelope detector
has been widely used to compare the EMG-tension relationship, especially if the tension is chan-
ging with time. If constant tension experiments are done, it is sufficient to calculate the average of
the full-wave rectified signal, which is the same as that derived from a long time-constant linear
envelope circuit. Both linear and nonlinear relationships between EMG amplitude and tension
have been discovered. Typical of the work reporting linear relationships is an early study by
Lippold (1952) on the calf muscles of humans. Zuniga and Simons (1969) and Vredenbregt
and Rau (1973), on the other hand, found quite nonlinear relationships between tension and
EMG in the elbow flexors over a wide range of joint angles. Both these studies were, in effect,
static calibrations of the muscle under certain length conditions; a reproduction of their results is
presented in Figure 9.19.
Another way of representing the level of EMG activity is to count the action potentials over a
given period of time. Close et al. (1960) showed a linear relationship between the count rate and
the IEMG, so it was not surprising that the count rate increased with muscle tension in an almost
linear fashion.
The relationship between force and linear envelope EMG also holds during dynamic changes
of tension. Inman et al. (1952) first demonstrated this by a series of force transducer signals that
were closely matched by the envelope EMG from the muscle generating the isometric force.
Gottlieb and Agarwal (1971) mathematically modeled this relationship with a second-order
low-pass system (e.g. a low-pass filter). Under dynamic contraction conditions, the tension is
seen to lag behind the EMG signal, as shown in Figure 9.20. The delay is caused by the fact that
the twitch corresponding to each m.u.a.p. reaches its peak 40–100 ms afterward. Thus, as each
motor unit is recruited, the resulting summation of twitch forces will also have a similar delay
behind the EMG.
(a) During a gradual build-up and rapid relaxation.

(b) During a short 400-ms contraction.
56° 138° 162°

t1
t2
Calf muscle integrated EMG
1.5
Elbow flexor EMG (mv)
1.0
0.5
0
Tension 0 100 200 300 400
(from Lippold, 1952) Wrist force (N)
(from Vredenbregt and Rau, 1973)
Figure 9.19 Relationship between the average amplitude of EMG and the tension in the muscle in isometric
contraction. Linear relationships have been shown by some researchers, while others report the EMG amplitude
to increase more rapidly than the tension.
ISTUDY
(a)
2s
(b)
400 ms
Figure 9.20 EMG and muscle tension recordings during varying isometric contractions of the biceps muscles.
Note the delay between the EMG and the initial build-up of tension, time to reach maximum tension, and drop
of tension after the EMG has ceased.
In spite of the reasonably reproducible relationships, the question still arises as to how valid
these relationships are for dynamic conditions when many muscles act across the same joint:
1. How does the relationship change with length? Does the length change merely alter the
mechanical advantage of the muscle, or does the changing overlap of the muscle fibers
affect the EMG itself (Vredenbregt and Rau, 1973)?
2. How do other agonist muscles share the load at that joint, especially if some of the muscles
have more than one function (Vredenbregt and Rau, 1973)?
3. In many movements, there is antagonist activity. How much does this alter the force being
predicted by creating an extra unknown force?
With the present state of knowledge, it appears that a suitably calibrated linear envelope EMG
can be used as a coarse predictor of muscle tension for muscles whose length is not changing
rapidly.
9.4.2 Electromyogram During Muscle Shortening and Lengthening

In order for a muscle to do positive or negative work, it must also undergo length changes while it
is creating tension. Thus, it is important to see how well the EMG can predict tension under these
more realistic conditions. A major study has been reported by Komi (1973). In it, the subject did
both positive and negative work on an isokinetic muscle-testing machine. The subject was asked
to generate maximum tension while the muscle lengthened or shortened at controlled velocities.
The basic finding was that the EMG amplitude remained
ISTUDY
9.4 RELATIONSHIP BETWEEN ELECTROMYOGRAM AND BIOMECHANICAL VARIABLES 211
tension during shortening and increased tension during eccentric contractions (see Section 8.2.1).
Such results support the theory that the EMG amplitude indicates the state of activation of the
contractile element, which is quite different from the tension recorded at the tendon. Also, these
results combined with later results (Komi et al., 1987) indicate that the EMG amplitude associ-
ated with negative work is considerably less than that associated with the same amount of positive
work. Thus, if the EMG amplitude is a relative measure of muscle metabolism, such a finding
supports the experiments that found negative work to have somewhat less metabolic cost than
positive work.
9.4.3 Electromyogram Changes During Fatigue

Muscle fatigue occurs when the muscle tissue cannot supply the metabolism at the contractile
element, because of either ischemia (insufficient oxygen) or local depletion of any of the meta-
bolic substrates. Mechanically, fatigue manifests itself by decreased tension, assuming that the
muscle activation remains constant, as indicated by a constant surface EMG or stimulation rate.
Conversely, the maintenance of a constant tension after onset of fatigue requires increased motor
unit recruitment of new motor units to compensate for a decreased firing rate of the already
recruited units (Vredenbregt and Rau, 1973). Such findings also indicate that all or some of
the motor units are decreasing their peak twitch tensions but are also increasing their contraction
times. The net result of these changes is a decrease in tension.
Fatigue not only reduces the muscle force but also may alter the shape of the motor action
potentials. It is not possible to see the changes of shape of the individual m.u.a.p.’s in a heavy
voluntary contraction. However, an autocorrelation shows an increase in the average duration of
the recruited m.u.a.p. (see Section 9.1.4). Also, the EMG spectrum shifts to reflect these duration
changes; Kadefors et al. (1973) found that the higher-frequency components decreased. The net
result is a decrease in the EMG frequency spectrum, which has been attributed to the following:
1. Lower conduction velocity of the action potentials along the muscle fibers below the non-
fatigued velocity of 4.5 m/s (Mortimer et al., 1970; Krogh-Lund and Jørgensen, 1991).
2. Some of the larger and faster motor units with shorter duration m.u.a.p.’s dropping out.
3. A tendency for the motor units to fire synchronously, which increases the amplitude of the
EMG. Normally, each motor unit fires independently of others in the same muscle so that
the EMG can be considered to be the summation of a number of randomly timed m.u.a.p.’s.
However, during fatigue, a tremor (8–10 Hz) is evident in the EMG and force records.
These fluctuations are neurological in origin and are caused by motor units firing in syn-
chronized bursts.
The major EMG measures of fatigue are an increase in the amplitude of the EMG and a
decrease in its frequency spectrum. One of those measures is median power frequency, fm, which
is the frequency of the power spectral density function below which half the power lies and above
which the other half of the power lies:
fm ∞ ∞
2 2 1
X f df = X f df = X 2 f df (9.9)
2
0 fm 0
where X(f) is the amplitude of the harmonic at frequency f, and X2(f) is the power at frequency f.
A representative paper that shows the majority of these measures (Krogh-Lund and Jørgensen,
1991) reports a 10% decrease in conduction velocity, a 45% decrease in the median power fre-
quency, and a 250% increase in the root mean square (rms) amplitude. The median power fre-
quency has also been shown to be highly correlated with
ISTUDY
bandwidth (15–45 Hz) and the ratio of the high-frequency bandwidth (>95 Hz) to the low-
frequency bandwidth (Allison and Fujiwara, 2002). An alternate and very common statistical
measure (Öberg et al., 1994) is MPF:
f X 2 f df
0
MPF = F
Hz (9.10)
X 2 f df
0
where F is the maximum frequency analyzed.

There is virtually no difference in these two frequency measures and their correlations with
independent measures of fatigue (Kerr and Callaghan, 1999).
9.5 REFERENCES
Allison, G. T. and T. Fujiwara. “The Relationship between EMG Median Frequency and Low Frequency
Band Amplitude Changes at Different Levels of Muscle Capacity,” Clin. Biomech. 17: 464–469, 2002.
Andreassen, S. and A. Rosenfalck. “Relationship of Intracellular and Extracellular Action Potentials of Skel-
etal Muscle Fibers,” CRC Crit. Rev. Bioeng. 6: 267–306, 1981.
Basmajian, J. V. “Electrodes and Electrode Connectors,” in New Developments in Electromyography and
Clinical Neurophysiology, Vol. 1, J. E. Desmedt, Ed. (Karger, Basel, Switzerland, 1973), pp. 502–510.
Bouisset, S. “EMG and Muscle Force in Normal Motor Activities,” in New Developments in Electromyog-
raphy and Clinical Neurophysiology, Vol. 1, J. E. Desmedt, Ed. (Karger, Basel, Switzerland, 1973),
pp. 547–583.
Boyd, D. C., P. D. Lawrence, and P. J. A. Bratty. “On Modelling the Single Motor Unit Action Potential,”
IEEE Trans. Biomed. Eng. BME-25: 236–242, 1978.
Buchthal, F., C. Guld, and P. Rosenfalck. “Propagation Velocity in Electrically Activated Fibers in Man,”
Acta Physiol. Scand. 34: 75–89, 1955.
Buchthal, F., F. Erminio, and P. Rosenfalck. “Motor Unit Territory in Different Human Muscles,” Acta
Physiol. Scand. 45: 72–87, 1959.
Calvert, T. W. and A. B. Chapman. “Relationship Between Surface EMG and Force Transients in Muscle:
Simulation and Experimental Results,” Proc. IEEE 65: 682–689, 1977.
Close, J. R., E. D. Nickel, and A. B. Todd. “Motor Unit Action Potential Counts,” J. Bone Joint Surg. 42-A:
1207–1222, 1960.
Crosby, P. A. “Use of Surface Electromyography as a Measure of Dynamic Force in Human Limb Muscles,”
Med. Biol. Eng. Comput. 16: 519–524, 1978.
DeLuca, C. J. “Precision Decomposition of EMG Signals,” Methods Clin. Neurophysiol. 4: 1–28, 1993.
DeLuca, C. J. “The Use of Surface Electromyography in Biomechanics,” J. Appl. Biomech. 13:
135–163, 1997.
DeLuca, C. J., R. S. LeFever, M. P. McCue, and A. P. Xenakis. “Control Scheme Governing Concurrently
Active Motor Units during Voluntary Contractions,” J. Physiol. 329: 129–142, 1982.
Erim, Z., C. J. DeLuca, K. Mineo, and T. Aoki. “Rank-Ordered Regulation of Motor Units,” Muscle Nerve 19:
563–573, 1996.
Forssberg, H. “Ontogeny of Human Locomotor Control,” Exp. Brain Res. 57: 480–493, 1985.
Fuglevand, A. J., D. A. Winter, A. E. Patla, and D. Stashuk. “Detection of Motor Unit Action Potentials with
Surface Electrodes: Influence of Electrode Size and Spacing,”
ISTUDY
9.5 REFERENCES 213
Gersten, J. W., F. S. Cenkovich, and G. D. Jones. “Harmonic Analysis of Normal and Abnormal
Electromyograms,” Am. J. Phys. Med. 44: 235–240, 1965.
Gottlieb, G. L. and G. C. Agarwal. “Dynamic Relationship between Isometric Muscle Tension and the
Electromyogram in Man,” J. Appl. Physiol. 30: 345–351, 1971.
Hermens, H. J., B. Freriks, C. Disselhorst-Klug, and G. Rau. “Development of Recommendations for SEMG
Sensors and Sensor Placement Procedures,” J. Electromyogr. Kinesiol. 10: 361–374, 2000.
Inman, V. T., H. J. Ralston, J. B. Saunders, B. Feinstein, and E. W. Wright. “Relation of Human Electromy-
ogram to Muscular Tension,” Electroencephalogr. Clin. Neurophysiol. 4: 187–194, 1952.
Johansson, S., L. E. Larsson, and R. Ortengren. “An Automated Method for the Frequency Analysis of Myo-
electric Signals Evaluated by an Investigation of the Spectral Changes Following Strong Sustained
Contractions,” Med. Biol. Eng. 8: 257–264, 1970.
Kadaba, M. P., M. E. Wootten, J. Gainey, and G. V. B. Cochran. “Repeatability of Phasic Muscle
Activity: Performance of Surface and Intramuscular Wire Electrodes in Gait Analysis,” J. Orthop. Res.
3: 350–359, 1985.
Kadefors, R. “Myo-electric Signal Processing as an Estimation Problem,” in New Developments in Electro-
myography and Clinical Neurophysiology, Vol. 1, J. E. Desmedt, Ed. (Karger, Basel, Switzerland, 1973),
pp. 519–532.
Kadefors, R., I. Petersen, and H. Broman. “Spectral Analysis of Events in the Electromyogram,” in New
Developments in Electromyography and Clinical Neurophysiology, Vol. 1, J. E. Desmedt, Ed. (Karger,
Basel, Switzerland, 1973), pp. 628–637.
Kerr, D. and J. P. Callaghan “Establishing a Relationship Between Spectral Indicators of Fatigue and Ratings
of Perceived Discomfort,” Proc. 31st Conf. Human Factors Assoc. Canada, 301–307, 1999.
Komi, P. V. “Relationship between Muscle Tension, EMG, and Velocity of Contraction under Concentric
and Eccentric Work,” in New Developments in Electromyography and Clinical Neurophysiology,
Vol. 1, J. E. Desmedt, Ed. (Karger, Basel, Switzerland, 1973), pp. 596–606.
Komi, P. V. and E. R. Buskirk. “Reproducibility of Electromyographic Measures with Inserted Wire Elec-
trodes and Surface Electrodes,” Electromyography 10: 357–367, 1970.
Komi, P. V., M. Kaneko, and O. Aura. “EMG Activity of the Leg Extensor Muscles with Special Reference to
Mechanical Efficiency in Concentric and Eccentric Exercise,” Int. J. Sports Med. 8: 22–29, Suppl., 1987.
Krogh-Lund, C. and K. Jørgensen. “Changes in Conduction Velocity, Median Frequency, and Root-Mean-
Square Amplitude of the Electromyogram During 25% Maximal Voluntary Contraction of the Triceps
Brachii Muscle, to Limit of Endurance,” Eur. J. Physiol. 63: 60–69, 1991.
Kwatny, E., D. H. Thomas, and H. G. Kwatny. “An Application of Signal Processing Techniques to the Study
of Myoelectric Signals,” IEEE Trans. Biomed. Eng. BME-17: 303–312, 1970.
Lippold, O. C. J. “The Relationship Between Integrated Action Potentials in a Human Muscle and its Isomet-
ric Tension,” J. Physiol. 177: 492–499, 1952.
de Lorente, No, R. “A Study of Nerve Physiology: Analysis of the Distribution of Action Currents of Nerve in
Volume Conductors,” Stud. Rockefeller Inst. Med. Res. 132: 384–477, 1947.
McGill, K. C., L. J. Dorfman, J. E. Howard, and E. V. Valaines “Decomposition Analysis of the Surface
Electromyogram,” Proc. 9th Ann. Conf. IEEE Eng. Med. Biol. Soc., 2001–2003, 1987.
Milner-Brown, H. S. and R. B. Stein. “The Relation between Surface Electromyogram and Muscular Force,”
J. Physiol. 246: 549–569, 1975.
Milner-Brown, H. S., R. B. Stein, and R. Yemm. “Contractile Properties of Human Motor Units During Vol-
untary Isometric Contractions,” J. Physiol. 228: 285–306, 1973.
Mortimer, J. T., R. Magnusson, and I. Petersen. “Conduction Velocity in Ischemic Muscle: Effect on EMG
Frequency Spectrum,” Am. J. Physiol. 219: 1324–1329, 1970.
Murray, M. P., G. B. Spurr, S. B. Sepic, G. M. Gardner, and L. A. Mollinger. “Treadmill vs. Floor Walking:
Kinematics, Electromyogram and Heart Rate,” J. Appl. Physiol. 59: 87–91, 1985.
Öberg, T., L. Sandsjö, and R. Kadefors. “EMG Mean Power Frequency: Obtaining a Reference Value,” Clin.
Biomech. 9: 253–257, 1994.
ISTUDY
Person, R. S. and L. N. Mishin. “Auto and Crosscorrelation Analysis of the Electrical Activity of Muscles,”
Med. Biol. Eng. 2: 155–159, 1964.
Plonsey, R. “Volume Conductor Fields of Action Currents,” Biophys. J. 4: 317–328, 1964.
Plonsey, R. “The Active Fiber in a Volume Conductor,” IEEE Trans. Biomed. Eng. BME-21: 371–381, 1974.
Rosenfalck, P. “Intra and Extracellular Potential Fields of Active Nerve and Muscle Fibres,” Acta Physiol.
Scand. 321: 1–165, Suppl., 1969.
Vredenbregt, J. and G. Rau. “Surface Electromyography in Relation to Force, Muscle Length and Endurance,”
in New Developments in Electromyography and Clinical Neurophysiology, Vol. 1, J. E. Desmedt, Ed.
(Karger, Basel, Switzerland, 1973), pp. 607–622.
Winter, D. A., G. Rau, R. Kadefors, R. Broman, and C. DeLuca “Units, Terms and Standards in Reporting
EMG Research,” Report of an AD Hoc Committee of the Internat. Soc. Electrophysiol, Kinesiol.
August 1980.
Winter, D. A., A. J. Fuglevand, and A. J. Archer. “Crosstalk in Surface Electromyography: Theoretical and
Practical Estimates,” J. Electromyogr. Kinesiol. 4: 15–26, 1994.
Zuniga, E. N. and D. G. Simons. “Non-linear Relationship between Averaged Electromyogram Potential and
Muscle Tension in Normal Subjects,” Arch. Phys. Med. 50: 613–620, 1969.
ISTUDY
10
MODELING OF HUMAN MOVEMENT
BRIAN A. KNARR1, TODD J. LEUTZINGER1, AND NAMWOONG KIM2
1
Department of Biomechanics, College of Education, Health, and Human Sciences, University of Nebraska
at Omaha, Omaha, NE, USA
2
Division of Health & Kinesiology, Incheon National University, Incheon, Republic of Korea
10.0 INTRODUCTION
The analysis of human movement has been expanding and improving for several decades.
Previously, joint kinetics during human movement have been analyzed through inverse dynamics
solutions. Detailed in Chapters 5 and 7, inverse dynamics solutions utilize the body’s measured
kinematics and external forces applied to the body (i.e. ground reaction forces) to estimate inter-
nal joint reaction forces and joint moments. A link-segment model uses previously calculated
joint reaction forces and moments to predict the forces that caused a given movement. However,
inverse dynamics is the opposite of how our bodies actually move in real life. The true sequence
that occurs in our bodies begins with the central nervous system (CNS) utilizing a top-down
approach sending signals to the muscles through the peripheral nervous system. If the signal
to the muscles is strong enough, this results in the recruitment of agonist and antagonist muscles.
The net effect of muscle forces results in a moment acting about that joint. This moment leads to
the acceleration of adjacent body segments and ultimately causes a person to move.
Computer modeling that implements the top-down approach utilized by our bodies is referred
to as a forward solution. Unfortunately, the necessary information and constraints required to
perform a forward solution are considerably greater than performing an inverse dynamics solu-
tion. For example, if we wish to calculate the ankle and knee moments on one limb utilizing the
inverse dynamics approach, we only need data from the adjacent segments involved in the joint
moment calculation (in this case, the foot and the shank of the respective leg). On the other hand,
to calculate a forward solution, we must model all parts of the body including the thighs, hips, and
trunk. This is a direct result of the interlimb coupling of forces that occurs in the body. Any given
joint moment acts directly on the two adjacent segments that make up the joint. A sort of ripple
effect occurs and these two segments create reaction forces on segments further away from the
original joint. In human gait, the plantarflexor muscles create a plantarflexion moment about the
ankle. Based on the magnitude of the plantarflexion moment, the ankle will plantarflex at a

Companion website:
ISTUDY
216 MODELING OF HUMAN MOVEMENT
particular speed and with a certain amount of torque. Due to the resulting accelerations experi-
enced, the horizontal and vertical reaction forces at the knee will be altered. This in turn has a
direct effect on the acceleration of the thigh which will alter the reaction forces at the hip, con-
tralateral hip, and trunk. In this way, the accelerations of multiple body segments are affected by
the plantar flexor moment.
Not only must we model each segment of the body during a forward solution, but it is crucial
that our model be anatomically correct. If we have errors in our anatomical model, even if it is not
directly related to the joint in question, we will have errors in our model predictions. For example,
if a body segment has a disproportionately inaccurate mass, a joint has a missing constraint, or the
constraints placed on a joint are unrealistic, the entire link system will start to generate displace-
ment errors and these errors will continue to multiply over time at increasing rates. Therefore,
despite the fact forward solutions utilize a top-down approach similar to how our bodies interact
with the world, accurate anatomical information is crucial when running forward simulations
given they present more opportunities for errors to be introduced.
10.1 REVIEW OF FORWARD SOLUTION MODELS
Human locomotion and movement mechanics have been the most intriguing topics for research-
ers performing modeling simulations. Unfortunately, due to the complexities of both human
movement and the link-segment model required to perform these simulations, biomechanical
simplifications have been made and many of the simulations were confined to short periods
of the movement. The first true forward simulation model is known as the inverted pendulum
model (Cavagna et al., 1963). This model used a simple inverted pendulum illustration to
describe the movement of the body’s center of mass throughout the gait cycle. In 1980, McMahon
and Mochon made further additions to the inverted pendulum model that utilized the law of con-
servation of energy (Mochon and McMahon, 1980). The model used the assumptions that the
body exchanges kinetic energy for potential energy based on stance phase and swing phase
and the vertical displacement of the body’s center of mass. This model is known as the ballistic
model. Hemami proposed a three-segment three-dimensional model with rigid legs and no feet
(Hemami, 1980), and Pandy and Berme simulated single support only, using a five-segment pla-
nar model with no feet (Pandy and Berme, 1988). Unfortunately, such serious simplifications
could not produce valid answers. Even with more complete models, many researchers con-
strained parts of their models kinematically by assuming sinusoidal trajectories of the trunk or
pelvic segments (Beckett and Chang, 1968; Chao and Rim, 1973; Townsend, 1981). Such con-
straints violate one of the major requirements of a true simulation. Onyshko and Winter modeled
the body as a seven-segment system (two feet, two legs, two thighs, and a head, arm, and trunk
(HAT) segment), but the model did not satisfy the requirements of internal validity because of
certain anatomical constraints (sagittal plane movement only at all three joints plus a rigid foot
segment) (Onyshko and Winter, 1980). The Onyshko and Winter model was eventually made to
walk but only by altering the moment patterns. Such results should alert researchers that two
wrongs can make a right, but this does not necessarily make it correct. An incomplete model will
result in a valid movement only if faulty motor patterns are used.
In terms of modeling simulations, simple motions have been modeled and deemed reasonably
successfully. Phillips et al. modeled the swinging action of a human limb using the accelerations
and moments of the hip during the swing phase of running (Phillips et al., 1983). Hemami et al.
modeled the sway of the body in the coronal plane with locked out knees using adductor/abductor
actuators at the hip and ankle as input to define the stability of the total system (Hemami
et al., 1982).
More recent modeling of three-dimensional (3D) gait has been somewhat more successful.
One of the major problem points in previous attempts was the modeling of initial foot contact
with the ground. As a solution, springs were used to
ISTUDY
10.1 REVIEW OF FORWARD SOLUTION MODELS 217
demonstrated by the bottom of the human foot and shoe. Unfortunately, the use of these springs
resulted in extremely large accelerations of the foot segment and similarly large spikes in the
ground reaction forces. A viscoelastic model of the foot with an array of parallel springs and dam-
pers under the rigid foot segment was used to address this issue (Gilchrist and Winter, 1996).
In 1997, a 3D nine-segment model of walking that used ADAMS software (discussed in
Section 8.2) demonstrated some success using the inverse dynamic joint moments as inputs
(Gilchrist and Winter, 1997). Nonlinear springs at the knee, ankle, and metatarsal phalangeal
joints constrained those joints to their anatomical range. Linear springs were used at the hips,
and dampers were used at all joints to ensure a smooth motion. However, it became apparent
that any small error in the joint moments after about 500 ms resulted in accumulating kinematic
errors that ultimately became too large. This resulted in the model either becoming unbalanced or
ultimately collapsing. The build-up of these errors is an inherent characteristic of any forward
solution. The double integration of the segment accelerations caused by the input and reaction
moments and forces result in displacement errors that increase over time and can only be cor-
rected by continuous fine-tuning of the input joint moments. Thus, we are forced to violate some
of the constraints of forward solution modeling as listed in Section 10.1.1.
10.1.1 Assumptions and Constraints of Forward Solution Models

1. The link-segment model has the same assumptions as those presented in Section 5.0.1 for
the inverse dynamic solutions.
2. No kinematic constraints should be applied to the model whatsoever. The model must be
permitted to fall over, jump, or collapse as dictated by the motor inputs.
3. The initial conditions must include the position and velocity of every model segment.
4. The only inputs to the model are externally applied forces and internally generated muscle
forces or moments.
5. The model must incorporate all important degrees of freedom and constraints.
6. External reaction forces must be calculated. For example, the ground reaction forces would
be equal to the algebraic summation of the mass-acceleration products of all segments when
the feet are on the ground. Partitioning of the reaction forces when two or more points of the
body are in contact with external objects is a separate and possibly major problem.
10.1.2 Potential of Forward Solution Simulations

The sky is the limit in regards to the potential for groundbreaking research and discovery when
utilizing simulations. Unfortunately, due to the severe constraints associated with forward
dynamics, the full potential of using simulations has not been realized. However, before we
can use simulations for research, we must first make sure the link-segment representation of
our model is anatomically correct and valid as indicated in Section 10.1.1. Researchers have been
working on testing the internal validity of forward simulation models, but more work is still
needed to ensure the validity of simulation results (Hicks et al., 2015; Fregly et al., 2012; Taylor
et al., 2017).
Traditionally, testing a model’s internal validity requires an inverse dynamics solution to cal-
culate the moments at each joint in each plane of movement (sagittal, frontal, and coronal). Then,
by using these motor patterns as inputs along with measured initial conditions, the forward solu-
tion should reproduce the originally measured kinematics from the inverse dynamics solution. If
the model does not pass this test, it is counterproductive to use the model to answer any func-
tionally related questions. All that will result will be an erroneous set of motor patterns overlaid
on an erroneous biomechanical model. Fortunately, recent advancements in the field of instru-
mented prostheses have greatly progressed the validation process for musculoskeletal models and
will be highlighted in later sections of this chapter.
ISTUDY
10.2 MUSCLE-ACTUATED SIMULATION OF MOVEMENT
The human body is an extremely complex dynamic system composed of ligaments, tendons,
bones, muscles, and other soft tissues that interact with the world around it. All these components
within the body work together to generate forces and create movement. Muscle-actuated simula-
tions provide the opportunity to study all these factors together and allow for a better understand-
ing of human movement.
There are two main outcomes of musculoskeletal simulations that aid in our understanding of
human movement: The first is the estimation of internal joint forces during dynamic movements.
Previously, internal joint forces were only able to be measured through in vivo measurements.
Additionally, if surrogate measures were used, such as an inverse dynamics approach, muscle
forces were left unaccounted for. Through the use of musculoskeletal simulations, we can
now estimate forces occurring inside the joint while adequately accounting for the muscle forces
that contribute to this load. The second main outcome is the capability of performing “what if”
studies to establish cause–effect relationships for phenomena that occur in the body. Traditional
methods used to analyze human movement, such as motion capture, have been limited mainly to
joint angles, moments, and powers. However, with musculoskeletal simulation, we can reason-
ably estimate the forces occurring inside a joint (Knarr and Higginson, 2015), determine the
effects of surgery prior to the procedure (Homayouni et al., 2015), analyze landing mechanics
on surfaces deemed unsafe for human subjects (DeMers et al., 2017), investigate the effectiveness
of assistive devices (Rosenberg and Steele, 2017), and so much more.
In this chapter, we will discuss the key components of musculoskeletal simulations to better
understand how this tool can be used to answer a variety of different research questions.
10.2.1 Musculoskeletal Modeling

Developing a musculoskeletal model is the first step needed to investigate the underlying
mechanisms responsible for movement disorders that would otherwise be impossible due to
the complexity of the human body. The basic component for building a musculoskeletal model
is the geometry of the musculoskeletal system. Musculoskeletal geometry includes skeletal
geometry, joint structure, structure of passive objects, muscle attachment sites, and muscle
moment arms.
10.2.1.1 Skeletal Geometry. Accurate skeletal geometry is essential in creating a valid and
reliable musculoskeletal model. Inaccurate skeletal geometry can lead to the false representation
of muscle origins, insertion points, and muscle lines of action, which ultimately will lead to inac-
curate estimations of muscle forces. Individual subject-specific skeletal geometry is often
neglected due to the considerable investment of time and effort that is required.
The initial steps in building the geometry of a musculoskeletal model begin with the construc-
tion of individual rigid bodies that make up the skeletal system of the model. Synonymous to the
human body, one rigid body (parent body) is connected to and able to manipulate another rigid
body (child body) through the connection at a joint. A joint is an articulation between two rigid
bodies in the model. In a model, a joint can allow anywhere from zero degrees of freedom to
six degrees of freedom. Unfortunately, there is no perfect joint definition for a model that
describes the human joint. For example, the knee joint is often defined as a hinge joint with
one degree of freedom permitted in the sagittal plane. However, in the human body, the knee
joint is able to translate anteriorly/posteriorly, abduct/adduct, and internally/externally rotate
as well. Because these other movements are not as prevalent as flexion/extension in the knee
joint, they are often omitted. The level of complexity of a joint in a model largely varies depend-
ing on the model and the research question being asked. If a model is too complex it becomes
difficult to validate and computationally expensive to
ISTUDY
10.2 MUSCLE-ACTUATED SIMULATION OF MOVEMENT 219
10.2.1.2 Passive Structures (Ligament, Cartilage, Fascia, etc.). Skeletal muscles are the pri-
mary actuators, driving forces in a model that act as muscles, that generate joint motion in the
body. However, passive structures such as ligament, cartilage, and fascia have mechanical prop-
erties that influence muscle force generation and can be influential toward the forces and
moments occurring at a joint during dynamic movements. For example, when running, the plan-
tarflexor muscles are the primary drivers that allow an individual to move their legs and produce
power. However, elastic energy stored and released in the Achilles tendon during certain phases
of the gait cycle plays a significant role in the ability to run. It is believed a major reason why
kangaroos are able to run so fast is because of their Achilles tendon (Morgan et al., 1978). Addi-
tionally, tendon slack length (resting tendon length) is an important parameter included in the
musculoskeletal model. Because the tendon acts as nonlinear spring, tendon slack length substan-
tially affects muscle force production both in timing and magnitude.
10.2.1.3 Muscle Moment Arms. A muscle generating force vector at a perpendicular distance
from a joint center of rotation creates a moment about that joint.
T=r×F (10.1)
In Equation 10.1, the moment about a joint is the result of the vector cross product between the
force of the muscle crossing that joint (F) and the muscle moment arm (r). The muscle moment
arm is the perpendicular distance between the joint center of rotation and the muscle force vector.
The muscle moment arm carries just as much weight in the calculation of joint moments as the
muscle force itself. Therefore, it is important to measure the moment arm about a joint accurately.
The use of magnetic resonance imaging (MRI) imaging is helpful in measuring this perpendicular
distance between the muscle force vector and the joint center of rotation, but it is important to note
this measurement cannot just be taken once. The moment arm must be the perpendicular distance
which means it changes as the joint angle changes. Therefore, to accurately calculate the moment
about a joint, one needs to accurately measure the length of the moment arm at each point of the
joint’s range of motion. The use of an MRI machine to perform this task would be extremely
tedious and expensive which begs the question “is there a better way?”.
Another way to define the moment arm is the tendon excursion definition. This definition
allows for the computational discovery of a moment arm with respect to the joint angle without
having to measure it “by hand” with an MRI. By simultaneously measuring the position of a
muscle-tendon unit and the joint angle, one can then compute the derivative of the muscle-tendon
unit length with respect to joint angle. This results in the ability to calculate the muscle moment
arm as a function of joint angle throughout the range of motion of the joint (An, 2007). A much
simpler way than taking MRIs. Using this definition, the moment arm of a muscle about a joint is
a function of both the joint angle and muscle-tendon length. Both of which can be affected by the
origin and insertion of the muscle, i.e. the muscle attachment sites making this an important factor
to accurately represent within the model.
10.2.1.4 Muscle Attachment Sites. As discussed previously, the moment created about a spe-
cific joint is dependent on the moment arm. That being said, it is important to accurately place the
origin and insertion point for each muscle-tendon unit when building a musculoskeletal model. In
a musculoskeletal model, muscles can be attached to bone using a variety of different methods
depending on the type of muscle, its location on the body, and its overall range of motion. The
first and simplest form of creating muscle attachments is through the use of fixed points. When
using fixed points, the muscle path is set in a straight line and is defined by a series of attachment
points that are fixed to the body. The points where the muscle attaches to the bone are called fixed
points. The second form of muscle attachment used in musculoskeletal modeling is called
via points. The use of via points is similar to using fixed
ISTUDY
are fixed to the body. However, they differ from fixed points because they are only fixed at the
origin and insertion points of the muscle. This allows via points to be used in simple cases of
wrapping when a joint moves to a certain angle and can assist with accurately portraying the mus-
cle’s line of action. For example, via points can be used on the ulna when the elbow flexes beyond
a specific angle. Wrap points are attachment points that are automatically calculated by OpenSim
to traverse over the surface of wrap objects. This is done to constrain the path of the muscle and
guide it to avoid interference with other geometric structures in the model. Lastly, moving muscle
points are used when the muscle paths are supposed to move with the joint and occur when the
XYZ offsets in a body act as functions of the joint motion rather than a constant as seen with fixed
points.
10.2.1.5 Muscle-Tendon Parameters. Skeletal muscles have certain mechanical parameters

that are common across all muscles. Specifically, in musculoskeletal modeling, there are four
important parameters that are included in the Hill-type muscle model: optimal fiber length, pen-
nation angle at optimal fiber length, maximum isometric force, and maximum contraction veloc-
ity. The Hill-type muscle model has been widely used and is one of the common and basic muscle
models. Optimal fiber length is the sarcomere length that can generate maximum isometric force
within the muscle. Optimal muscle fiber length can be determined using the following equation
assuming that all sarcomeres have the same length in a muscle fiber.
O = n ∗ lO
lM S
(10.2)
In Equation (10.2), lM S
O is the optimal fiber length, lO is the optimal sarcomere length, and n is
the number of sarcomeres in series. Optimal muscle fiber length affects the force–length and
force–velocity curve, which in turn affects muscle fiber force production.
The angle of the fibers within a muscle also plays a role in the force production of a muscle.
This is known as the pennation angle and is the angle at which the muscle fibers are arranged
within the muscle (see Chapter 8 for more detail on pennation angles of muscle fibers). The
way a muscle contracts, the number of fibers in a muscle, and the amount of force a muscle
can produce are all affected by its pennation angle.
An isometric contraction occurs when the muscle generates force but the joint angle does not
change and the length of the muscle remains the same. A muscle’s maximum isometric force is
the maximum amount of force generated from a muscle when the muscle fibers are at their opti-
mal length and do not change. This maximum isometric force is a critical measurement in terms of
measuring a muscle’s ultimate force production. Unfortunately, maximum isometric force pro-
duction for each individual muscle is difficult to obtain because it is hard to truly isolate a single
muscle due to the body’s redundancy. Therefore, physiological cross-sectional area (PCSA) is a
viable alternative that is often used to estimate a muscle’s maximum isometric force. It should be
noted however, a muscle’s anatomical cross-sectional area and its PCSA are not always the same
thing. The PCSA of a muscle is the cross-section of the muscle taken perpendicular to the muscle
fibers. Therefore, if we take a muscle’s PCSA and multiply it by its specific tension (the force per
unit area of a muscle) we are able to calculate the maximum isometric force for any muscle.
Values for the specific tension of a muscle can vary and should be considered when developing
a model (Dickerson et al., 2007; Buchanan et al., 1998; Arnold et al., 2010).
Muscle force production is also influenced by muscle contraction velocity. Each muscle con-
tains two different muscle fiber types: fast-twitch and slow-twitch fibers. Slow-twitch muscle
fibers (also called type 1) are more aerobic in nature, less resistant to fatigue, and generate force
at a slow rate. Conversely, fast-twitch muscle fibers (type 2) are anaerobic in nature, fatigue
quickly, and generate force at fast rates. All skeletal muscle in the body contains both fast
and slow-twitch muscle fibers, but the ratio of fibers depends on each individual muscle and even
each individual person. For example, a sprinter will have a
ISTUDY
fibers compared to a marathon runner who will mainly have muscle built up of slow-twitch fibers.
Training can alter fiber type within a muscle to a certain degree but this make-up of fast-twitch
versus slow-twitch muscle fibers is largely determined in the first few years of life and can be
highly genetic (Kenney et al., 2015). The muscle contraction velocity of a muscle is important
as it influences the force–velocity profile. According to the force–velocity relationship for mus-
cles, the slower a concentric contraction occurs, the more force the muscle is able to produce (see
Chapter 8 for muscular force–velocity relationships). Although fast-twitch muscles may be able
to generate force at higher speeds, they are unable to optimize their force output at these higher
speeds. Therefore, muscle contraction velocity and muscle fiber type are important to consider
when modeling the musculoskeletal system.
Tendons behave similarly to nonlinear springs and like skeletal muscle, they too have a force–
length relationship. The amount of force that can be transmitted from muscle to bone by a tendon
is dependent on the length of the tendon. The tendon force–length relationship is not linear and
the slope of the force–length curve is the stiffness of the tendon. The slope can be divided into
three different parts on the curve and represents three different areas where the tendon elicits three
distinctly different properties. In the first three percent of the curve when the tendon has not been
stretched, the tendon is very flexible but produces very little force. Similar to stretching a rubber
band to fling it across the room, if the rubber band has only been stretched very little it will go a
very short distance when released. In the second part of the curve, as the tendon is stretched more
and more, the amount of force that can be produced increases at a linear rate. There is a point
where you can increasingly stretch the rubber band and it will go incrementally further across
the room. Lastly, the last ten percent of the curve is where the tendon is pushed past its mechanical
limit and plastic deformation occurs resulting in injury to the tendon. This is similar to when the
rubber band becomes overstretched and is in danger of snapping. Therefore, in order to produce
maximal force from a muscle, it is crucial the tendon length is also optimal. This is crucial to
avoid injury while still stretching the tendon enough to produce ample power from the muscle.
This force–length relationship of the tendon emphasizes the importance of properly modeling
tendon properties when performing musculoskeletal simulations.
10.2.2 Control
Because the human musculoskeletal system is a redundant system (there are more unknown vari-
ables than equations), it is not well understood how the CNS controls the human body to adapt to
its ever-changing environment. For example, there are multiple ways to grab a coffee cup in front
of you, and it is assumed the CNS chooses the most effective movement based on certain criteria.
In this case, if your musculoskeletal model needs to produce a desired motion, what control strat-
egy should be used to produce muscle force for the desired movement?
The most common control models include (i) optimization approaches and (ii) EMG-driven
approaches. Optimization approaches are further divided into static optimization (SO) and
dynamic optimization.
10.2.2.1 Optimization.
10.2.2.1.1 Static Optimization of Muscle Forces As discussed in previous chapters, a net joint
moment can be estimated using an inverse dynamics approach from motion capture data. How-
ever, there are many plausible solutions that could possibly generate the same calculated net joint
moment. To solve this problem of redundancy, an optimization approach, utilizing certain basic
criteria (e.g. Minimizing or maximizing cost function) and constraints, can decompose the net
joint moment into individual muscle forces at a single instant in time. This decomposition pro-
vides greater insight into the forces occurring around a specific joint and removes the guesswork
in interpreting the cause of certain joint moments. Static equilibrium problems ignore time-
dependent characteristics of human movement and have
ISTUDY
modeling because it is computationally efficient and able to estimate muscle forces during
dynamic movement.
SO has been used for a variety of different clinical initiatives to help advance the field of reha-
bilitative care. For example, Steele et al. utilized SO to investigate the change in muscle forces
that occur in individuals with cerebral palsy who suffer from crouch gait (Steele et al., 2012). This
estimation of muscle forces allowed researchers to investigate knee joint contact forces present in
this population as many individuals with crouch gait also struggle with knee pain. Additionally,
Arnold et al. investigated muscle activations of the lower extremities during walking to rank the
potential of each muscle to alter lower joints (Arnold et al., 2005). This was done in an attempt to
determine the cause of crouch gait. Wesseling et al. utilized two different SO techniques to inves-
tigate how different estimates of muscle force would affect the calculation of hip joint contact
forces (Wesseling et al., 2015). Rosenberg and Steele performed simulations to investigate
how powered and passive ankle-foot orthoses would affect muscle demand and recruitment in
individuals with cerebral palsy (Rosenberg and Steele, 2017). Lastly, Kian et al. used SO and
musculoskeletal modeling to estimate the muscle and joint forces occurring in the glenohumeral
joint and compare their findings to an EMG-driven model of identical structure (Kian et al.,
2019). These are just a few examples of how SO is used in the literature to investigate muscle
forces and improve the world of clinical biomechanics.
10.2.2.1.2 Dynamic Optimization Dynamic optimization is fundamentally distinguished

from SO for two reasons. First, dynamic optimization employs forward dynamics. Therefore,
the dynamic optimization approach often calculates muscle excitation patterns to control the
musculoskeletal model. Then, these muscle excitation patterns are used to estimate muscle forces,
moments, joint angular accelerations, joint angular velocities, and joint angles to ultimately
calculate the positions of the musculoskeletal model. Second, the goal of motor tasks can be
included in the formulation of the problem for the whole simulation of human movement.
Because of the time-dependent characteristic of forward simulation, we can calculate time his-
tories of muscle forces for the whole movement. This is distinguishably different from SO that
calculates muscle forces at a single instant in time. One of the downfalls associated with SO is the
abrupt activation and deactivation of muscles. Because of the time independence of SO, the esti-
mation of muscle forces can be quite different for each time step. In other words, SO chooses a
different set of muscle groups and a different magnitude of muscle forces for each time step
resulting in a lack of continuity for muscle activations throughout a movement. Contrary to
SO, dynamic optimization is designed to determine muscle excitation in conjunction with for-
ward dynamics to find optimal motion. For example, we can find a set of muscle excitation pat-
terns that maximize jump height or minimize energy expenditure during walking with additional
weight. This is advantageous because it allows for the prediction of complex human movements
without the need for experimental data.
Dynamic optimization often requires more detailed musculoskeletal models and computation
time compared to SO. To decrease computational cost, hybrid optimization approaches such as
computed muscle control (CMC) and direct collocation have been used with tracking experimen-
tal data. CMC is an algorithm to solve tracking problems with forward dynamics along with SO.
Kinematics and muscle excitation levels from SO are used as feedback to drive forward dynamic
simulations that track experimental data. By doing so, CMC can significantly reduce computation
time. However, CMC involving SO still solves an optimization problem at each single time point,
which may limit the benefits of dynamic optimization.
Direct collocation, a recently developed method for dynamic optimization, converts dynamic
equations of motion into algebraic constraints. Both the controls (muscle excitations) and the
states (positions and velocities) are parameterized while solving an optimization problem result-
ing in efficient computation. OpenSim Moco (Musculoskeletal Optimal Control) provides a way
for researchers to utilize direct collocation to answer their scientific questions (Dembia et al.,
2020). With computationally efficient direct collocation and
ISTUDY
to solve various kinds of problems including motion tracking problems, predictive dynamic opti-
mization, EMG-driven simulation, model parameter optimization, and device design.
Similar to SO, dynamic optimization has been used in the literature to help advance the field of
clinical biomechanics. For example, Davy and Audu utilized a dynamic optimization solution to
predict muscle forces in the swing phase of gait (Davy and Audu, 1987). Additionally, McLean
et al. investigated knee forces in the sagittal plane to determine if these shear forces could rupture
the anterior cruciate ligament (ACL) during sidestepping to aid in athlete injury prevention
(McLean et al., 2004). Kuzelicki et al. used a dynamic optimization technique to compute the
trajectories of a sit-to-stand movement in amputees as a step in the development of assistive
devices to aid in the sit-to-stand transition for this population (Kuzelicki et al., 2005). Eskinazi
and Fregly used a dynamic optimization approach to calculate muscle forces, joint contact forces,
and body motion simultaneously, to speed up the computation process and help facilitate model-
based efforts to improve surgical outcomes and rehabilitative interventions (Eskinazi and Fregly,
2018). These are just a few examples of how dynamic optimization solutions are used to
investigate human movement and improve the lives of many from a clinical perspective.
10.2.3 OpenSim
One key component in the field of dynamic simulations and musculoskeletal modeling is Open-
Sim simulation software (Delp et al., 2007). OpenSim is an open-source software that allows
users to both develop musculoskeletal models and run dynamic simulations for a variety of dif-
ferent movements. OpenSim encourages collaboration between labs and institutions as a way for
the scientific community to build an archive of simulations that can be exchanged, verified, ana-
lyzed, and improved by multiple different researchers. Previously, researchers spent considerable
effort replicating simulations from other labs because the software used by each lab was custo-
mized or simply unavailable. By being open source, OpenSim has sped up the development and
sharing of musculoskeletal models and simulation technology tremendously and since has
allowed the field of biomechanics to make leaps and bounds in the optimization of musculoskel-
etal simulations.
10.2.3.1 Assumptions and Constraints. Like all models and dynamic simulations, there are
many assumptions and constraints with the OpenSim software. Each individual model has its
own constraints and assumptions that dictate its motion. For example, many of the joints within
an OpenSim model are constrained to reduced degrees of freedom compared to the real world
with some models even constraining joints to no motion at all. Additionally, all models have lim-
itations or make assumptions which can cause dynamic inconsistencies in the simulations. Some
common assumptions and limitations include models not accounting for ligaments, tendons,
cartilage, and other soft tissues; models assuming separate tendons exist for each muscle;
oversimplified geometry; and the ability for muscles to actually go through bones. Additionally,
OpenSim assumes muscles to be massless and disregards muscle fatigue during motion. Fiber
type distribution within a muscle is often ignored and fiber lengths and velocities are assumed
to be constant throughout the muscle. These limitations, assumptions, and constraints within a
model can create inconsistencies in the simulated movement and should be made known to
the reader indicating the potential for inaccuracies in the final simulated movement.
10.2.3.2 How It Works. OpenSim follows a logical progression that begins with creating and
scaling a model and advances to finding a set of muscle excitations that properly fit the exper-
imental motion performed in the lab (Figure 10.1).
10.2.3.3 Scaling. First, each model is scaled to the anthropometrics of the study participant
by utilizing the relationship between the marker locations of the experimental motion capture
data and the virtual marker locations on the OpenSim
ISTUDY
Experimental
GRF
1. Scale 2. IK 3. ID
Subject-specific
Joint angles
model
Experimental
Joint angles
GRF
Fast muscle
activation/force Static
estimation optimization
5.1. SO Rigid tendon
Adjusted torso
COM
4. RRA
Joint angles
Muscle
excitations/force Static
estimation optimization
5.2. CMC Tendon Forward
compliance dynamics
Figure 10.1 General musculoskeletal modeling workflow. ∗IK, inverse kinematics; ID, inverse dynamics; SO,
static optimization; RRA, residual reduction algorithm; CMC, computed muscle control.
frame and mass center locations along with the force application and muscle attachment points.
This is all based on scale factors that are computed in the scaling process. Next, the mass and
inertial properties of each body segment are scaled proportionally based on the mass of the
participant using generalized ratios from human or cadaveric work. Unfortunately, these data
often come from single cadaver studies or cadavers of older individuals making them an
inaccurate representation of healthy young individuals. Additionally, individual properties
of these mass ratios can vary a great deal across age, sex, and disease state making this a lim-
itation within the scaling process. After scaling the mass and inertial properties of each body
segment, muscles and other model components that are dependent on length such as ligaments
and muscle actuators are scaled. A ratio of the length of the model component before scaling to
the length of the model component after scaling is used as the scale factor to scale the
component’s length-dependent properties. Lastly, the final step is to move the virtual marker
locations of the model to match the marker locations of the motion capture data. Once the vir-
tual marker locations of the model are aligned with the experimental marker data, the model is
considered to be scaled. Unfortunately, the virtual and experimental markers will never align
perfectly. OpenSim’s best practices for marker errors should be consulted to assess simulation
quality.
ISTUDY
10.2.3.4 Inverse Kinematics. Once the model has been scaled to match the study participant
anthropometrics, an inverse kinematics (IK) solution is performed to determine the joint angles of
the model that most accurately reproduce the experimental kinematics from the motion capture
data. This is done through a least-squares calculation in which the difference between the exper-
imental marker data and the virtual marker data from the model are minimized throughout the
motion of the simulation.
10.2.3.5 Residual Reduction Algorithm. Once the virtual markers are adequately aligned
with the experimental motion capture data, the motion goes through the residual reduction
algorithm. The residual reduction algorithm, commonly known as RRA, seeks to make the
joint angles and translations of the model dynamically consistent with the experimental
motion performed in the lab. This is done by altering the mass center of a model’s torso
to make the motion dynamically consistent with the ground reaction force data of the move-
ment. By adjusting the mass of the body segments in the model and making small perturba-
tions to the motion trajectories, OpenSim is able to decrease the inconsistency between the
experimental motion and the model. This will result in lower “hand of god” residual forces
required for the simulation to run. The residuals are calculated and averaged throughout the
entire movement providing suggestions for adjustments to both the model translations and
body mass segments. A performance standard is used to calculate tracking errors across the
joint angles and is repeated at each instant in time throughout the simulation until the end of
the desired movement.
10.2.3.6 Static Optimization. The fourth step of the simulation progression involves solving
for the model’s muscle activations along with other forces that may be responsible for the stud-
ied movement pattern. This is done using one of two optimization techniques: CMC or SO. SO
is an extension of inverse dynamics, and utilizes the known positions, velocities, and accelera-
tions of a model to find the muscle activations and other forces responsible for the given
motion. This is done using a performance criterion that minimizes the sum of the muscle
activations.
n
p
J= am (10.3)
m=1
In Equation 10.3, n is the number of muscles in the model, am is the activation level of a given
muscle m at a particular point in time and p is the user-defined constraint.
10.2.3.7 Computed Muscle Control. On the other hand, CMC calculates coordinated muscle
excitations that drive the simulated movement while tracking the desired kinematics through a
forward dynamics solution. CMC first calculates the desired accelerations that will push the
model coordinates to be more like the experimental movement. Then CMC calculates the actuator
controls that distribute forces across muscles in an attempt to achieve the desired accelerations
that will match the tracked kinematics. This process is performed for each instant in time through-
out the simulation. Lastly, the CMC algorithm uses these actuator controls to produce a forward
dynamic simulation of the experimental movement. This process is repeated at each instant in
time throughout the simulation until the end of the desired movement. The benefit of using
CMC is that it is computationally efficient like SO while performing a harmonious forward
dynamic solution.
Upon the completion of CMC, a variety of analysis plug-ins can be performed on the model
that are able to calculate joint contact forces, muscle-induced accelerations, and muscle power
among others.
ISTUDY
10.2.3.7.1 Review of Literature Since its inception, OpenSim has been used by a variety of
different researchers to investigate a wide range of outcomes through a variety of different meth-
odologies. For example, both SO and CMC are used to estimate muscle activations and muscle
forces. Both are available for investigators to utilize and both have their advantages and disad-
vantages, but it is important to know the differences between the two methods before using them
in a simulation. Previous literature has investigated the differences between SO and CMC and
found CMC often overestimates muscle activations and muscle forces because it accounts for
passive muscle dynamics and activation-contraction dynamics. However, a study by Roelker
et al. investigated differences in muscle activation between SO and CMC using two different
models and found that the agreement between EMG readings and SO versus CMC was both mus-
cle-dependent and model-dependent rather than a fixed characteristic of the optimization tech-
nique (Roelker et al., 2020). Regardless of which method is used, both allow for the
investigation of a variety of different outcome variables such as muscle activation dynamics,
mass center accelerations, joint contact forces, and more. For example, Simpson et al. investi-
gated the feasible muscle activation ranges for a variety of different muscles during a full gait
cycle (Simpson et al., 2015). Hamner et al. used induced acceleration analyses to investigate indi-
vidual muscle contributions to the center of mass accelerations while running at different speeds
(Hamner et al., 2010). Lastly, Knarr et al. investigated the knee joint contact forces that occurred
during walking utilizing four different methods varying in the model’s specificity to the study
participant (Knarr and Higginson, 2015). As demonstrated here, OpenSim has a wide variety
of applications and although we mention the general linear progression one takes when using
OpenSim, it is not uncommon for researchers to expand on that line to continue to progress
the field forward.
10.2.3.7.2 Model Applications One of the main outcomes the creators of OpenSim desired
was for collaboration across labs to help improve the field of musculoskeletal modeling. In
an attempt to promote collaboration and field advancement, the creators of OpenSim spon-
sored “the grand knee challenge” at the 2010 ASME Summer Bioengineering Conference.
The grand knee challenge datasets were collected from individuals with an instrumented
knee implants (Fregly et al., 2012). The goal of this challenge was to encourage scientists
to develop musculoskeletal models that would accurately predict muscle forces and ulti-
mately knee joint contact forces. The CAMS knee data set is another example of using
instrumented implants to better understand knee joint function. The CAMS knee data set
was developed to better understand how internal movement of the knee relates to internal
knee joint loading (Taylor et al., 2017).
Additionally, OpenSim has been used for a variety of different clinical applications. For exam-
ple, Homayouni et al. utilized OpenSim to investigate passive mechanical implants used to
improve an individual’s ability to grasp and pinch objects following hand tendon-transfer surgery
(Homayouni et al., 2015). Additionally, muscle-actuated forward dynamic simulations allow for
the prediction of underlying muscular function, which drives an individual’s movement. This is
critical when investigating knee joint contact forces given muscle forces are a significant contrib-
utor to knee joint loading. Using modeling techniques, Sasaki and coworker demonstrated the
vastii muscles are the primary contributors to the axial knee joint contact force in early stance
with additional contributions from the rectus femoris, and biceps femoris short head (Sasaki and
Neptune, 2010). The gluteus maximus also contributed to the axial knee joint contact force during
early stance despite it not crossing the knee joint through its contribution to the ground reaction
force vector. During late stance, the triceps surae muscles were the main contributors to the axial
forces occurring at the knee. Shelburne et al. investigated the contributions of muscles, ligaments,
and ground reaction forces to medial and lateral knee joint loading (Shelburne et al., 2006). The
authors demonstrated that medial compartment loading of the tibiofemoral joint was mainly
attributed to the orientation of the ground reaction force vector whereas the quadriceps and
ISTUDY
gastrocnemius muscles provided a substantial contribution to lateral joint loading. Lastly, Myers
et al. utilized OpenSim to investigate the effect of strengthening the hip abductors following a total
hip arthroplasty on joint contact forces at the hip, knee, ankle, and low back (Myers et al., 2019).
These are just a few examples of how OpenSim has been used to investigate human movement and
improve our understanding of the human body to improve the lives of many from a clinical
perspective.
10.2.4 EMG-Driven Modeling

Electromyography, or EMG, has been widely used in biomechanics research to measure muscle
activations and investigate CNS control. For this reason, EMG data have been used as a control
signal to drive musculoskeletal models. Specifically, EMG data have been used to generate mus-
cle-tendon unit excitations to drive forward dynamic solutions. If you recall, a forward dynamic
simulation begins with a neural command, which is comparable to how our bodies perform move-
ment tasks in the real world. When using EMG-driven modeling, this neural command is repre-
sented as the amplitude of the EMG signal (Figure 10.2). The EMG signal is then processed to
represent the magnitude of the muscle activation patterns. Once we have solved for the muscle
activation, we must determine the muscle forces that result from these activations through the use
of a muscle model. Many muscle models that focus on characterizing the physiological aspects of
muscle are very complex and become computationally expensive when modeling forces in mul-
tiple muscles. The Hill-type muscle model (referred to as phenomenological) is simpler than most
other models such as the Huxley (Huxley, 1957) and distributed moment models (Gielen et al.,
2000) because it focuses on modeling the function of the muscle and is less concerned with the
actual muscle physiology. Because of this, the Hill-type muscle model is often used with EMG-
driven modeling. A Hill-type muscle model places the contractile element of the muscle fibers in
series with the viscoelastic component of the tendon to estimate the force generated by the mus-
cle-tendon unit. However, as mentioned previously, each muscle has its own optimal fiber length,
maximum contraction velocity, pennation angle, and other factors which dictate the muscle’s
ability to generate force effectively. Therefore, it is important to account for these specific muscle
parameters within the Hill-type model when modeling muscle forces. Unfortunately, the forces
produced by muscle-tendon units are nonlinear due to the changes in muscle fiber length as the
muscle contracts. Therefore, one must find the muscle-tendon force at each point of a time series
to accurately measure muscle forces throughout the duration of the movement being modeled.
As discussed previously, muscle-tendon unit anatomy plays a vital role in ensuring accurate
calculations of the moment created about a joint. Therefore, it is crucial to accurately determine
these variables before they are inserted in the model. The muscle-tendon moment arm is directly
related to the calculation for determining joint moments, but can be difficult to find because it is
constantly changing with the angle of the joint. However, if we know the muscle-tendon unit’s
length and the angle of the joint, the moment arm can be found rather easily using the tendon
displacement method (An, 2007). Once muscle-tendon unit forces and moment arms are calcu-
lated, we can calculate joint moments through the multiplication of force and the moment arm. If
we know these two parameters for every muscle, we can calculate the net moment about a joint.
Models can be validated through the use of a nonlinear optimization technique that aims to reduce
the squared error between the model predicted joint moment and the measured joint moment.
Using this technique, model coefficients can be refined and tuned for individual participants.
However, one should be wary of allowing too many variables to be altered because models con-
taining a lot of varying parameters tend to be inaccurate and display poor predictive power for
novel data. Therefore, it is a delicate balance between having a model that accurately replicates
the anatomy of the human body without having too many parameters that need to be adjusted at
each step of a time series. Similar to the Hill-type model, musculoskeletal models should accu-
rately describe the human body without being overly complex.
ISTUDY
Muscle activation Musculotendon kinematics
Joint
EMG angles
Musculotendon
activation
Musculo
length
Musculotendon
moment arms
Musculotendon dynamics
CE
PE
θ
adjustment
Parameter
Musculotendon
force
Moment calculation
Evaluation
calibration
Experimental
moment
Figure 10.2 Schematic depiction of the EMG-driven modeling.
An EMG-driven modeling approach has been used extensively in the field of biomechanics to
provide insights for injury prevention, motor control, surgical intervention, and pathological dis-
orders. One strength the EMG-driven modeling approach has over the optimization approach is
that EMG-driven modeling can account for individual CNS control rather than adopting a stan-
dardized performance criterion. Additionally, the EMG-driven approach is computationally more
efficient than dynamic optimization because this approach does not require the estimation of mus-
cle excitations through dynamic optimization. However, EMG-driven modeling has some
ISTUDY
limitations. One limitation is that muscular activity from deep musculature is not measurable and
cannot be accounted for in this approach. Indwelling EMG can measure deep muscle activity, but
it is invasive and can alter one’s natural behavior during a dynamic task. Additionally, inherent
technical challenges raise the question as to whether EMG data can truly represent CNS control.
However, the EMG-driven modeling approach appears to be more suitable for individuals who
have a disability or history of injury as their movement strategies may be adapted to accommo-
date their disabilities or previous injuries. Lastly, EMG-driven modeling currently requires cal-
ibration procedures to determine physiological parameters such as tendon slack length, optimal
fiber length, EMG-to-activation recursive filter coefficients, maximum isometric force, and non-
linear shape factor (Lloyd and Besier, 2003; Buchanan et al., 2004; Manal and Buchanan, 2013).
The calibrated parameters need to be applied to a musculoskeletal model to predict muscle forces
using the EMG-driven modeling approach.
EMG-driven modeling is very beneficial for understanding muscle dynamics during move-
ment, but unfortunately, it can become overly complex when multiple segments are involved
making the modeling procedure computationally expensive. A hybrid method is available that
combines both forward and inverse dynamic solutions. The hybrid scheme is beneficial because
it allows for a cross-validation measure of the joint moments calculated from the forward dynam-
ics method within the error of the inverse dynamics method.
10.2.4.1 Model Applications. Previous researchers have utilized EMG-driven modeling to

investigate the mechanics of the human body through the use of a forward dynamics solution
(McGill and Norman, 1986; McGill, 1992; Cholewicki et al., 1995; van Dieën and Visser,
1999). Unfortunately, to perform a forward dynamics solution the estimation of muscle activa-
tions from EMG signals and the transformation of muscle activation to muscle force is not a
trivial task. The use of cost functions can be utilized to bypass the task of determining muscle
force from EMG data, but the question of “which cost function to use?” is highly debated. Addi-
tionally, it does not seem wise to replace the entire CNS with a simple, unverified equation. How-
ever, in the last several decades, EMG-driven neuromusculoskeletal modeling (Buchanan et al.,
2004) has been used to investigate human body function from the body’s neural signals. For
example, Kumar et al. utilized EMG-driven modeling to estimate muscle forces and joint loading
in individuals with knee osteoarthritis (Kumar et al., 2012). Knee joint loading is altered during
the onset of knee osteoarthritis and throughout its progression. Additionally, muscle forces are an
important factor in determining knee joint loading. EMG-driven modeling allowed Kumar et al.
to utilize subject-specific muscle activation patterns which provided greater accuracy in their esti-
mation of muscle forces and ultimately provided a better estimation of the loads occurring at the
knee joint (Kumar et al., 2012). Shao et al. utilized an EMG-driven musculoskeletal model to
estimate the muscle forces and joint moments occurring about the ankle in individuals who suf-
fered from a stroke (Shao et al., 2009). Especially during complex tasks like walking, muscle
activation and movement patterns are significantly altered in individuals post stroke. By utilizing
an EMG-driven musculoskeletal model, Shao et al. sought to better understand these changes and
develop improved rehabilitation techniques for this population. Manal et al. developed a real-time
EMG-driven model of the ankle with the intention of monitoring muscle forces during rehabil-
itation of injuries such as an Achilles tendon tear (Manal et al., 2011). Achilles tendon tears are
the kiss of death for many athletes, and it is important that patients and rehabilitation specialists
alike are able to see the muscle forces produced by the ankle to determine if the patient is working
hard enough to produce improvement, but not too hard that they would risk reinjury. The authors
believe this model is an integral step in improving the rehabilitation process for Achilles tendon
injuries allowing for athletes and laypersons alike to return to function sooner. Nikooyan et al.
developed an EMG-driven model of the shoulder that was able to account for co-contractions that
occur at the shoulder (Nikooyan et al., 2012). By developing an EMG-driven musculoskeletal
ISTUDY
model of this caliber, more reliable measurements of the forces occurring at the shoulder joint can
be obtained. This is especially crucial for individuals who have rotator cuff tears. Lastly, Koo and
Mak demonstrated the ability to use an EMG-driven model to predict individual muscle forces at
the elbow during dynamic movements (Koo and Mak, 2005). EMG-driven modeling truly gives
researchers greater insight into the muscle forces occurring at a variety of different joints and
helps increase the knowledge and understanding of the movements at these joints to ultimately
help improve the quality of life for a variety of different clinical populations.
10.3 MODEL VALIDATION
Musculoskeletal modeling is an excellent way to test hypotheses that are not safe or feasible for
human subjects to perform and many great scientific discoveries have come from modeling and
simulations. However, we must admit that there is no perfect model or control algorithm to
represent all variables related to human movement and physiological characteristics. Therefore,
before using a model and performing a simulation it is important to verify the model to ensure it is
valid, accurate, and reliable. According to the American Society of Mechanical Engineers
(ASME), validation of a model is “the process of determining the degree to which a model is
an accurate representation of the real world from the perspective of the intended uses of the
model” and verification of a model is “the process of determining that a computational model
accurately represents the underlying mathematical model and its solution” (Hicks et al., 2015).
The process of verifying and validating a model takes a variety of different steps.
1. Devise a research question that a modeling simulation can answer. In this step, it is impor-
tant to formulate a research question that will not only advance the field forward, but it is
also imperative to make sure modeling and simulations are both necessary and capable of
answering the research question.
2. Design methods for the study and create a plan for model validation and verification to add
credibility to the results.
3. Before any data is collected, it is important to verify the software that is being used to
ensure the model and algorithms used in the simulation are correct.
4. Compare the model used and validate the study results to other experiments in the literature.
5. Test the robustness of the model. This is often done by simulating other scenarios not tested
in the experiment.
6. Document the modeling and simulation methods used in the study and sharing the model
and simulations with other researchers to allow them to test the model.
7. Generate other hypotheses that do not need modeling simulations but are testable in the
real world.
One of the most common validation approaches used in the current literature is to compare
simulated muscle activations to muscle activations from EMG sensors. It is important to compare
temporal agreement between simulated muscle activation and muscle activations from EMG sen-
sors because muscle activation timing significantly impacts muscle force production. When com-
paring simulated muscle activations to EMG data, it is important to consider electromechanical
delay from EMG recordings. Additionally, when filtering EMG data different cutoff frequencies
have been shown to alter the electromechanical delay recorded by the EMGs (Manal et al., 2002;
Lloyd and Besier, 2003). As a result of these findings, investigators need to be cognizant of the
filtration methods that should be used for their data based on the specific aims of the study (see
Chapter 2 for signal processing methods). Unfortunately, there is no established standard for vali-
dating a model using EMG data, but visual confirmation for temporal agreement between sim-
ulated muscle activation and EMG activation is often used.
ISTUDY
10.4 REFERENCES 231
It is important to understand that a good agreement between experimental data and simulated
outcomes does not always mean that the model is correct. In complex models and movements, it
is challenging to detect errors from various sources such as simplified models, control algorithms,
and imperfect experimental data in complex models and movements. Therefore, there is a risk
that a combination of errors may mask or offset errors making simulated results seemingly cor-
rect. To reduce uncertainty from a model, sensitivity analysis is often used to investigate the
impact of any particular parameters on key variables. Creating and validating models requires
extensive effort and time, researchers are encouraged to share their knowledge, models, and
report errors to accelerate musculoskeletal simulation studies.
10.4 REFERENCES
An, K.-N. “Tendon Excursion and Gliding: Clinical Impacts from Humble Concepts,” J. Biomech. 40(4):
713–718, 2007. https://doi.org/10.1016/j.jbiomech.2006.10.008.
Arnold, A. S., F. C. Anderson, M. G. Pandy, and S. L. Delp. “Muscular Contributions to Hip and Knee Exten-
sion During the Single Limb Stance Phase of Normal Gait: A Framework for Investigating the
Causes of Crouch Gait,” J. Biomech. 38(11): 2181–2189, 2005. https://doi.org/10.1016/J.
JBIOMECH.2004.09.036.
Arnold, E. M., S. R. Ward, R. L. Lieber, and S. L. Delp. “A Model of the Lower Limb for Analysis of Human
Movement,” Ann. Biomed. Eng. 38(2): 269–279, 2010. https://doi.org/10.1007/s10439-00998525.
Beckett, R. and K. Chang. “An Evaluation of the Kinematics of Gait by Minimum Energy,” J. Biomech. 1(2):
147–159, 1968. https://doi.org/10.1016/0021-9290(68)90017-1.
Buchanan, T. S., S. L. Delp, and J. A. Solbeck. “Muscular Resistance to Varus and Valgus Loads at the
Elbow,” J. Biomech. Eng. 120(5): 634–639, 1998. https://doi.org/10.1115/1.2834755.
Buchanan, T. S., D. G. Lloyd, K. Manal, and T. F. Besier. “Neuromusculoskeletal Modeling: Estimation of
Muscle Forces and Joint Moments and Movements from Measurements of Neural Command,” J. Appl.
Biomech. 20(4): 367–395, 2004. https://doi.org/10.1123/jab.20.4.367.
Cavagna, G. A., F. P. Saibene, and R. Margaria. “External Work in Walking,” J. Appl. Physiol. 18 (January):
1–9, 1963. https://doi.org/10.1152/jappl.1963.18.1.1.
Chao, E. Y. and K. Rim. “Application of Optimization Principles in Determining the Applied Moments in
Human Leg Joints During Gait,” J. Biomech. 6(5): 497–510, 1973. https://doi.org/10.1016/0021-9290
(73)90008-0.
Cholewicki, J., S. M. McGill, and R. W. Norman. “Comparison of Muscle Forces and Joint Load from an
Optimization and EMG Assisted Lumbar Spine Model: Towards Development of a Hybrid Approach,”
J. Biomech. 28(3): 321–331, 1995. https://doi.org/10.1016/0021-9290(94)00065-c.
Davy, D. T. and M. L. Audu. “A Dynamic Optimization Technique for Predicting Muscle Forces in the Swing
Phase of Gait,” J. Biomech. 20(2): 187–201, 1987. https://doi.org/10.1016/0021-9290(87)90310-1.
Delp, S. L., F. C. Anderson, A. S. Arnold, P. Loan, A. Habib, C. T. John, E. Guendelman, and D. G. Thelen.
“OpenSim: Open-Source Software to Create and Analyze Dynamic Simulations of Movement,” IEEE
Trans. Biomed. Eng. 54(11): 1940–1950, 2007. https://doi.org/10.1109/TBME.2007.901024.
Dembia, C. L., N. A. Bianco, A. Falisse, J. L. Hicks, and S. L. Delp. “OpenSim Moco: Musculoskeletal Opti-
mal Control,” PLoS Comput. Biol. 16(December), 2020. https://doi.org/10.1371/JOURNAL.
PCBI.1008493.
DeMers, M. S., J. L. Hicks, and S. L. Delp. “Preparatory Co-Activation of the Ankle Muscles May Prevent
Ankle Inversion Injuries,” J. Biomech. 52(February): 17–23, 2017. https://doi.org/10.1016/
j.jbiomech.2016.11.002.
Dickerson, C. R., D. B. Chaffin, and R. E. Hughes. “A Mathematical Musculoskeletal Shoulder Model for
Proactive Ergonomic Analysis,” Comput. Methods Biomech. Biomed. Eng. 10(6): 389–400, 2007. https://
doi.org/10.1080/10255840701592727.
ISTUDY
van Dieën, J. H. and B. Visser. “Estimating Net Lumbar Sagittal Plane Moments from EMG Data. The Valid-
ity of Calibration Procedures,” J. Electromyogr. Kinesiol. 9(5): 309–315, 1999. https://doi.org/10.1016/
s1050-6411(99)00004-8.
Eskinazi, I. and B. J. Fregly. “A Computational Framework for Simultaneous Estimation of Muscle and Joint
Contact Forces and Body Motion Using Optimization and Surrogate Modeling,” Med. Eng. Phys.
54(April): 56–64, 2018. https://doi.org/10.1016/j.medengphy.2018.02.002.
Fregly, B. J., T. F. Besier, D. G. Lloyd, S. L. Delp, S. A. Banks, M. G. Pandy, and D. D. D’Lima. “Grand
Challenge Competition to Predict in vivo Knee Loads,” J. Orthopaed. Res. 30(4): 503–513, 2012. https://
doi.org/10.1002/jor.22023.
Gielen, A. W. J., C. W. J. Oomens, P. H. M. Bovendeerd, T. Arts, and J. D. Janssen. “A Finite Element
Approach for Skeletal Muscle Using a Distributed Moment Model of Contraction,” Comput. Methods Bio-
mech. Biomed. Eng. 3(3): 231–244, 2000. https://doi.org/10.1080/10255840008915267.
Gilchrist, L. A. and D. A. Winter. “A Two-Part, Viscoelastic Foot Model for Use in Gait Simulations,” J.
Biomech. 29(6): 795–798, 1996. https://doi.org/10.1016/0021-9290(95)00141-7.
Gilchrist, L. A. and D. A. Winter. “A Multisegment Computer Simulation of Normal Human Gait,” IEEE
Trans. Rehabil. Eng. 5(4): 290–299, 1997. https://doi.org/10.1109/86.650281.
Hamner, S. R., A. Seth, and S. L. Delp. “Muscle Contributions to Propulsion and Support During Running,” J.
Biomech. 43(14): 2709–2716, 2010. https://doi.org/10.1016/j.jbiomech.2010.06.025.
Hemami, H. “A Feedback On-Off Model of Biped Dynamics,” IEEE Trans. Syst. Man Cybern. 10(7): 376–
383, 1980 https://ieeexplore.ieee.org/abstract/document/4308518/.
Hemami, H., M. J. Hines, R. E. Goddard, and B. Freidman. “Biped Sway in the Frontal Plane with Locked
Knees,” IEEE Trans. Syst. Man Cybern. SMC-12(4): 577–582, 1982. https://pascal-francis.inist.fr/vibad/
index.php?action=getRecordDetail&idt=PASCAL83X0157347.
Hicks, J. L., T. K. Uchida, A. Seth, A. Rajagopal, and S. L. Delp. “Is My Model Good Enough? Best Practices
for Verification and Validation of Musculoskeletal Models and Simulations of Movement,” J. Biomech.
Eng. 137(2): 020905, 2015. https://doi.org/10.1115/1.4029304.
Homayouni, T., K. N. Underwood, K. C. Beyer, E. R. Martin, C. H. Allan, and R. Balasubramanian. “Mod-
eling Implantable Passive Mechanisms for Modifying the Transmission of Forces and Movements
Between Muscle and Tendons,” IEEE Trans. Biomed. Eng. 62(9): 2208–2214, 2015. https://doi.org/
10.1109/TBME.2015.2419223.
Huxley, A. F. “Muscle Structure and Theories of Contraction,” Prog. Biophys. Biophys. Chem. 7(January):
255–318, 1957. https://doi.org/10.1016/S0096-4174(18)30128-8.
Kenney, W. L., J. Wilmore, and D. Costill. Physiology of Sport and Exercise, 6th edition, Vol. 6. (2015), p. 42.
https://books.google.com/books/about/Physiology_of_Sport_and_Exercise_6th_Edi.html?
id=OfjPoQEACAAJ.
Kian, A., C. Pizzolato, M. Halaki, K. Ginn, D. Lloyd, D. Reed, and D. Ackland. “Static Optimization
Underestimates Antagonist Muscle Activity at the Glenohumeral Joint: A Musculoskeletal Modeling
Study,” J. Biomech. 97(December): 109348, 2019. https://doi.org/10.1016/j.jbiomech.2019.109348.
Knarr, B. A. and J. S. Higginson. “Practical Approach to Subject-Specific Estimation of Knee Joint Contact
Force,” J. Biomech. 48(11): 2897–2902, 2015. https://doi.org/10.1016/j.jbiomech.2015.04.020.
Koo, T. K. K. and A. F. T. Mak. “Feasibility of Using EMG Driven Neuromusculoskeletal Model for Pre-
diction of Dynamic Movement of the Elbow,” J. Electromyogr. Kinesiol. 15(1): 12–26, 2005. https://doi.
org/10.1016/J.JELEKIN.2004.06.007.
Kumar, D., K. S. Rudolph, and K. T. Manal. “EMG-Driven Modeling Approach to Muscle Force and Joint
Load Estimations: Case Study in Knee Osteoarthritis,” J. Orthopaed. Res. 30(3): 377–383, 2012. https://
doi.org/10.1002/jor.21544.
Kuzelicki, J., M. Zefran, H. Burger, and T. Bajd. “Synthesis of Standing-up Trajectories Using Dynamic
Optimization,” Gait Posture 21(1): 1–11, 2005. https://doi.org/10.1016/j.gaitpost.2003.11.004.
Lloyd, D. G. and T. F. Besier. “An EMG-Driven Musculoskeletal Model to Estimate Muscle Forces and Knee
Joint Moments in Vivo,” J. Biomech. 36(6): 765–776, 2003. https://doi.org/10.1016/s0021-9290(03)
00010-1.
ISTUDY
10.4 REFERENCES 233
Manal, K. and T. S. Buchanan. “An Electromyogram-Driven Musculoskeletal Model of the Knee to Predict
in vivo Joint Contact Forces During Normal and Novel Gait Patterns,” J. Biomech. Eng. 135(2): 021014,
2013. https://doi.org/10.1115/1.4023457/371375.
Manal, K., R. V. Gonzalez, D. G. Lloyd, and T. S. Buchanan. “A Real-Time EMG-Driven Virtual Arm,”
Comput. Biol. Med. 32(1): 25–36, 2002. https://doi.org/10.1016/s0010-4825(01)00024-5.
Manal, K., K. Gravare-Silbernagel, and T. S. Buchanan. “A Real-Time EMG-Driven Musculoskeletal Model
of the Ankle,” Multibody Syst. Dyn. 28(1): 169–180, 2011. https://doi.org/10.1007/S11044-011-9285-4.
McGill, S. M. “A Myoelectrically Based Dynamic Three-Dimensional Model to Predict Loads on Lumbar
Spine Tissues During Lateral Bending,” J. Biomech. 25(4): 395–414, 1992. https://doi.org/10.1016/
0021-9290(92)90259-4.
McGill, S. M. and R. W. Norman. “Partitioning of the L4-L5 Dynamic Moment into Disc, Ligamentous, and
Muscular Components During Lifting,” Spine 11(7): 666–678, 1986. https://doi.org/10.1097/00007632-
198609000-00004.
McLean, S. G., X. Huang, S. Anne, and A. J. Van Den Bogert. “Sagittal Plane Biomechanics Cannot Injure
the ACL During Sidestep Cutting,” Clin. Biomech. 19(8): 828–838, 2004. https://doi.org/10.1016/
J.CLINBIOMECH.2004.06.006.
Mochon, S. and T. A. McMahon. “Ballistic Walking,” J. Biomech. 13(1): 49–57, 1980. https://doi.org/
10.1016/0021-9290(80)90007-x.
Morgan, D. L., U. Proske, and D. Warren. “Measurements of Muscle Stiffness and the Mechanism of Elastic
Storage of Energy in Hopping Kangaroos,” J. Physiol. 282(1): 253–261, 1978. https://doi.org/10.1113/
JPHYSIOL.1978.SP012461.
Myers, C. A., P. J. Laz, K. B. Shelburne, D. L. Judd, J. D. Winters, J. E. Stevens-Lapsley, and B. S. Davidson.
“Simulated Hip Abductor Strengthening Reduces Peak Joint Contact Forces in Patients with Total Hip
Arthroplasty,” J. Biomech. 93(August): 18–27, 2019. https://doi.org/10.1016/j.jbiomech.2019.06.003.
Nikooyan, A. A., H. E. J. Veeger, P. Westerhoff, B. Bolsterlee, F. Graichen, G. Bergmann, and F. C. T. van
der Helm. “An EMG-Driven Musculoskeletal Model of the Shoulder,” Hum. Mov. Sci. 31(2): 429–447,
2012. https://doi.org/10.1016/J.HUMOV.2011.08.006.
Onyshko, S. and D. A. Winter. “A Mathematical Model for the Dynamics of Human Locomotion,” J. Bio-
mech. 13(4): 361–368, 1980 https://doi.org/10.1016/0021-9290(80)90016-0.
Pandy, M. G. and N. Berme. “Synthesis of Human Walking: A Planar Model for Single Support,” J. Biomech.
21(12): 1053–1060, 1988. https://doi.org/10.1016/0021-9290(88)90251-5.
Phillips, S. J., E. M. Roberts, and T. C. Huang. “Quantification of Intersegmental Reactions During Rapid
Swing Motion,” J. Biomech. 16(6): 411–417, 1983. https://doi.org/10.1016/0021-9290(83)90073-8.
Roelker, S. A., E. J. Caruthers, R. K. Hall, N. C. Pelz, A. M. W. Chaudhari, and R. A. Siston. “Effects of
Optimization Technique on Simulated Muscle Activations and Forces,” J. Appl. Biomech. (July): 1–20,
2020. https://doi.org/10.1123/jab.2018-0332.
Rosenberg, M. and K. M. Steele. “Simulated Impacts of Ankle Foot Orthoses on Muscle Demand and Recruit-
ment in Typically Developing Children and Children with Cerebral Palsy and Crouch Gait,” PLoS ONE
12(7), 2017. https://doi.org/10.1371/JOURNAL.PONE.0180219.
Sasaki, K. and R. R. Neptune. “Individual Muscle Contributions to the Axial Knee Joint Contact Force During
Normal Walking,” J. Biomech. 43(14): 2780–2784, 2010. https://doi.org/10.1016/
j.jbiomech.2010.06.011.
Shao, Q., D. N. Bassett, K. Manal, and T. S. Buchanan. “An EMG-Driven Model to Estimate Muscle Forces
and Joint Moments in Stroke Patients,” Comput. Biol. Med. 39(12): 1083–1088, 2009. https://doi.org/
10.1016/j.compbiomed.2009.09.002.
Shelburne, K. B., M. R. Torry, and M. G. Pandy. “Contributions of Muscles, Ligaments, and the Ground-
Reaction Force to Tibiofemoral Joint Loading During Normal Gait,” J. Orthopaed. Res. 24(10): 1983–
1990, 2006. https://doi.org/10.1002/jor.20255.
Simpson, C. S., M. Hongchul Sohn, J. L. Allen, and L. H. Ting. “Feasible Muscle Activation Ranges Based on
Inverse Dynamics Analyses of Human Walking,” J. Biomech. 48(12): 2990–2997, 2015. https://doi.org/
10.1016/J.JBIOMECH.2015.07.037.
ISTUDY
Steele, K. M., M. S. DeMers, M. H. Schwartz, and S. L. Delp. “Compressive Tibiofemoral Force During
Crouch Gait,” Gait Posture 35(4): 556–560, 2012. https://doi.org/10.1016/J.GAITPOST.2011.11.023.
Taylor, W. R., P. Schütz, G. Bergmann, R. List, B. Postolka, M. Hitz, J. Dymke et al. “A Comprehensive
Assessment of the Musculoskeletal System: The CAMS-Knee Data Set,” J. Biomech. 65(December):
32–39, 2017. https://doi.org/10.1016/j.jbiomech.2017.09.022.
Townsend, M. A. “Dynamics and Coordination of Torso Motions in Human Locomotion,” J. Biomech.
14(11): 727–738, 1981. https://doi.org/10.1016/0021-9290(81)90029-4.
Wesseling, M., L. C. Derikx, F. de Groote, W. Bartels, C. Meyer, N. Verdonschot, and I. Jonkers. “Muscle
Optimization Techniques Impact the Magnitude of Calculated Hip Joint Contact Forces,” J. Orthopaed.
Res. 33(3): 430–438, 2015. https://doi.org/10.1002/JOR.22769.
ISTUDY
11
STATIC AND DYNAMIC BALANCE
1
2
3
11.0 INTRODUCTION
The complex interconnectivity of the neuromuscular system makes it extremely challenging to

interpret the function of a single variable at any single joint during the time course of any total
body movement. The various levels of integration were described in Section 1.2 where the first
three levels were involved with neuromuscular integration to generate the moment profile at a
given joint. However, the reason for this final motor profile is not evident until we look at the
movement task to see how the muscles at all joints contribute to the final goal. Also, during
any given movement, an individual group of muscles may also have more than one simultaneous
subtask to accomplish.
Biomechanics continues to evolve the methodologies and technologies used to measure and
analyze the total body in 3D space and, therefore, is capable of identifying total body movement
synergies either during normal everyday tasks or in response to perturbations, either internal or
external in origin. Some very quick examples will be referenced to illustrate some of these total
body analyses.
1. MacKinnon and Winter (1993) reported the total balance control in the frontal plane during
normal walking. The hip abductors responded proactively to the gravitational and inertial
forces acting on the head, arms, and trunk (HAT) segment to keep it nearly vertical during
single support.
2. Eng et al. (1992) identified the total body responses to total arm voluntary movements in
the sagittal plane: the hip, knee, and ankle moments responded in anticipation with appro-
priate polarities to the polarity of the shoulder moment. A flexor shoulder moment resulted
in a posterior postural response (hip extensors, knee flexors, and ankle plantarflexors),
while an extensor shoulder moment resulted in an anterior postural response (hip flexors,
knee extensors, and ankle dorsiflexors).
†
Deceased

Companion website:
ISTUDY
236 STATIC AND DYNAMIC BALANCE
3. Rietdyk et al. (1999) identified the total body balance recovery mechanisms from medio-
lateral external perturbations of the upper body during standing.
4. Scholz and Schöner (1999) established the uncontrolled manifold concept as a method to
examine the relationship between segmental coordination and stability of the bodies center
of mass in 3-D space during a sit-to-stand task. Yamagata et al. (2021) also demonstrated its
validity in predicting falls during walking in the elderly.
Human gait is a complex bipedal movement with many subtasks that must be simultaneously
satisfied and that are continuously changing over the stride period. These tasks may be comple-
mentary or competitive (Winter, 1991): muscles generate and absorb energy at the same time as
they are also responsible for the control of balance and vertical collapse of the body. This chapter
is presented to detail major examples of synergistic motor patterns and demonstrates how kinetic
and electromyographic (EMG) profiles aid in identifying these movement synergies.
The term synergy is defined here as muscles collaborating toward a common goal and there-
fore to identify such synergies, it is critical to clarify the goal and over what period of time the
muscle groups collaborate.
11.1 THE SUPPORT MOMENT SYNERGY
In Chapter 5, a detailed description of the three lower limb moments over stance was presented.
Also introduced was the concept of a total limb extensor pattern called the support moment
Ms = Mk − Ma − Mh (Winter, 1980). Considerably more information about this synergy becomes
evident when we analyze repeat trials on the same subject and analyze the considerable variability
at each of the joints (Winter, 1984, 1991). Figure 11.1 presents the ensemble-averaged profiles
for the same subject over nine days, where the subject was instructed to walk with her natural
cadence. Note that the convention for this plot has changed, with extensor moments at each joint
being plotted +ve, thus Ms = Mh + Mk + Ma.
The joint angle kinematics over these nine trials was very consistent: the rms standard devi-
ation over the stride period was 1.5 at the ankle, 1.9 at the knee, and 1.8 at the hip. The cadence
remained within 2%. As can be seen, Mh and Mk show negligible variability over swing but
considerable variability over the stance period (CVh = 68%, CVk = 60%), while Mh + Mk shows
considerably reduced variability (CVh + k = 21%). Also, Ms, which is the sum of all three
moments, has CVs = 20%. Thus, on a trial-to-trial basis, there is some “trading off” between
the hip and knee; on a given day, the hip becomes more extensor, while the knee becomes more
flexor, and vice versa on subsequent days. We can quantify these day-to-day interactions by a
covariance analysis. Figure 11.2 presents the variances and covariances for these repeat day-
to-day trials along with 10 repeat trials done on a second subject done minutes apart. The hip
and knee variances and covariances are related as follows with all units in (N m)2
σ 2hk = σ 2h + σ 2k − σ 2h + k (11.1)
where
σ 2h and σ 2k are the mean hip and knee variances over stance
σ 2+ k is the mean variance of the hip + knee profile over stance
σ 2hk is the mean covariance between hip and knee over stance
The term σ 2hk can be expressed as a percent of the maximum possible, which would be 100% if
σ 2h+k= 0, meaning that day-to-day changes in Mh and Mk were completely out of phase and were
ISTUDY
11.1 THE SUPPORT MOMENT SYNERGY 237
Moment of force (natural cadence, n = 9)
2
Subject: WM22 Support
CV = 20%
Ext
1
TO
0
1 Hip
TO CV = 68%
Ext
0
Moment/body mass (N. m/kg)
–1
Hip + Knee
1 CV = 21%
TO
0
Ext
–1
Knee
1
CV = 60%
TO
Ext
–1 TO = Toe off (forced to 60%)

CV = Coefficient of variation Mean
(during STANCE phase) + – 1 std.dev.
2
Plantar
1 Ankle
TO CV = 18%
0
0 20 40 60 80 100
% of stride
Figure 11.1 Ensemble averaged moment profiles for repeat trials from the same subject over nine days. The
support moment, Ms = Mh + Mk + Ma with extensor moment at each joint plotted +ve. The hip + knee moment,
Mh + k = Mh + Mk is plotted to demonstrate the dramatic reduction in variability because of day-to-day trade-offs
between these two joints. See the text for a detailed discussion. (Reproduced with permission from Winter (1991).)
canceling each other out completely. Thus, the maximum possible σ 2hk = σ 2h + σ 2k , and the %
covariance is:
COV = σ 2hk σ 2h + σ 2k × 100 (11.2)
As is evident from Figure 11.2 over the nine days, the σ 2hk = 19 1 N m2, which is 89% of the
maximum; even σ 2ak is 76% of the maximum. The trial-to-trial covariances are somewhat reduced,
but this is because the individual joint variances were drastically smaller over repeat trials
minutes apart compared to repeat trials over days.
11.1.1 Relationship Between Ms and the Vertical Ground Reaction Force

As the support moment represents a summation of the extensor moments at all three joints, it
quantifies how much the total limb is pushing away from the ground. The profile of Ms has
the characteristic “double-hump” shape seen in the vertical ground reaction force, Fy. To check
this out, a linear correlation was done between the averaged Ms and Fy profiles for three groups of
ISTUDY
Day-to-day Trial-to-trial
σ 2h = 15.92 σ 2h = 5.52
σ 2hk = 19.12 σ 2k+h = 6.92 σ 2hk = 6.52 σ 2k+h = 4.12

(89%) (72%)
σ 2k = 12.62 σ 2k = 5.42
σ 2ak = 14.32 σ 2a + k = 8.12 σ 2ak = 5.62 σ 2a + k = 5.72

(76%) (49%)
σ 2a = 10.52 σ 2a = 5.92
All units in N.m

σ 2hk
σ 2hk = σ 2h + σ 2k – σ 2h + k COV = × 100%
σ 2h + σ 2k
Figure 11.2 Variance and covariance analyses of the joint moments of the subject shown in Figure 11.1 plus a
similar analysis of 10 repeat trials of a second subject done minutes apart. There is a very high covariance between
the hip and knee moments (89%) and a moderate covariance between the knee and ankle moments (75%). The trial-
to-trial covariances are still moderately high but are reduced because the individual joint variances are considerably
smaller. See the text for detailed calculations. (Reproduced with permission from Winter (1991).)
Ms and Fy for natural cadence (N = 19)

12 1.2
10 1
8 0.8
6 0.6
4 0.4
}
Fy (nat, N/kg) 0.2
2
R = 0.97
Ms (nat, Nm/kg)
0 0
–2 –0.2
0 10 20 30 40 50 60
% stride
Figure 11.3 Average vertical ground reaction force, Fy, and the average support moment, Ms, for 19 adult subjects
walking with their natural cadence. The close similarity in these two profiles yielded r = 0.97 from a linear
regression. (Reproduced with permission from Winter (1991).)
walking adults (Winter, 1991): those walking their natural cadence, fast cadence (natural + 20),
and slow cadence (natural − 20). Figure 11.3 is a plot of Ms and Fy averaged profiles for the
19 subjects walking with their natural cadence. Here, we see an almost perfect match between
the two profiles, and this is reinforced by the
ISTUDY
11.2 MEDIAL/LATERAL AND ANTERIOR/POSTERIOR BALANCE IN STANDING 239
r = 0.95 and for 17 slow walkers, r = 0.90. Even the atypical moment patterns seen in pathological
gait yielded r = 0.96 for two walking trials for a 69-year-old female total knee replacement patient
and r = 0.92 for two trials for a 69-year-old male knee replacement patient.
11.2 MEDIAL/LATERAL AND ANTERIOR/POSTERIOR BALANCE

IN STANDING
11.2.1 Quiet Standing

Standing has been the subject of considerable research with posture and balance being the
dominant tasks in both medial–lateral (M/L) and anterior–posterior (A/P) directions (Horak
and Nashner, 1986; Winter et al., 1996; Gage et al., 2003). The primary outcome measures
are center-of-pressure (COP) and center-of-mass (COM), and it has been shown that as an
inverted pendulum model (see Chapter 5) that for sway angles of less than 8 the COP and
COM are related to the horizontal acceleration of the COM in either the A/P or M/L directions
(Winter et al., 1996):
COP − COM = − Ix Wd = − Kx (11.3)
where
I is the moment of inertia of the total body about the ankle in the direction of interest
x is the horizontal acceleration of the COM in the direction of interest
d is the vertical distance from the ankle joints to the COM
W is the total body weight above the ankle joints.
Thus, we can think of COP–COM as an error signal in the balance control system for control-
ling the horizontal acceleration of the COM, and it forces us to focus on how the central nervous
system (CNS) controls the COP to achieve a stable balance. In quiet standing in the A/P direction,
COP is controlled by the ankle dorsiflexors/plantorflexors (Horak and Nashner, 1986), whereas
in the M/L direction, COP is controlled by the hip abductors/adductors in what has been
described as a “load/unload” mechanism (Winter et al., 1996). Figure 11.4 summarizes this
mechanism during quiet standing and shows when two force platforms are used the right and
the left vertical ground reaction forces along with the M/L COP. Note that these vertical ground
reaction forces oscillate about 50% of body weight and the fluctuations about this 50% are vir-
tually equal in magnitude and also exactly out of phase. The M/L COP (cm) is a weighted average
of the left and right vertical reaction forces and with the convention shown is in phase with the
right vertical reaction force. The horizontal ground reaction forces are negligible; thus, from our
inverse dynamics, the left and right hip moments are respectively in phase with the left and right
vertical reaction forces. Thus, this “load/unload” mechanism is accomplished by hip abductor/
adductor moments that are exactly equal in magnitude and also 180 out of phase. A detailed
analysis of the COP and COM waveforms has attributed these motor patterns to a simple stiffness
control (Winter et al., 1998). This stiffness control is not reactive because the COP is virtually in
phase with the COM except the COP oscillates with larger amplitude on either side of the COM
and the COP–COM error signal [Equation (11.3)] keeps accelerating the COM toward a central
position. This mechanism can be considered as a bilateral synergy because the CNS has to keep
the left and right hip abductors/adductors at a small level of muscle tone so that the miniscule M/L
sway results in small synchronized fluctuations in hip frontal-plane moments sufficiently large to
maintain M/L balance. The fact that this balance mechanism does not involve continuous reactive
control allows all the reactive sensors to be on “standby” ready to react to an unexpected pertur-
bation. One final question relates to what causes the fluctuations in the COM. The COM in these
studies was estimated from a 14-segment total body model (see Chapter 4), which included four
trunk segments; this was necessary because of internal mass shifts mainly from the lungs and
heart (Winter and Lodge, 1966; Hunter and Kearney,
ISTUDY
Left vertical reaction force Right vertical reaction force

56 54
54 52
% body weight
% body weight
52 50
50 48
48 46
46 44
2 4 6 8 10 12 14 16 Net 2 4 6 8 10 12 14 16
Time (s) COP Time (s)
0.8
Copi Copr
0.4
Displacement (cm)
–0.4
2 4 6 8 10 12 14 16
–0.8 Time (s)
–1.2 M/L displacement of COPnet
Figure 11.4 Left and right vertical ground reaction forces from two force platforms from a subject standing
quietly. Note that the left and right vertical reaction forces oscillate about 50% of body weight and that these
oscillations are virtually equal in amplitude and exactly out of phase. The M/L COP is a weighted average of
these two reaction forces and with the convention shown is in phase with the right vertical reaction force. See
the text for details of this load/unload mechanism as a synergy to maintain balance during quiet standing.
(Reproduced with permission from Winter (1991).)
11.2.2 Medial Lateral Balance Control During Workplace Tasks

When standing on both feet human beings must maintain their balance whether they are standing
quietly or whether they are performing manual tasks with their upper extremities. A/P balance is
controlled by the plantarflexors/dorsiflexors (Horak and Nashner, 1986), whereas M/L balance is
controlled by the hip abductors/adductors (Winter et al., 1996). M/L balance was identified as a
“load/unload” mechanism, where increased abductor forces on one side are accompanied by
decreased abductor forces on the contralateral side. Such a pattern attempts to lift the pelvis/
HAT mass, thus loading the ipsilateral side, while simultaneously unloading the contralateral
side. This loading/unloading moves the COP toward the ipsilateral foot, thus causing an accel-
eration of the COM of the body toward the contralateral side.
This synergistic pattern was examined in an ergonomics task, where subjects stood in front of a
table carrying out a variety of manual tasks for a period of two hours (Nelson-Wong et al., 2008).
Surface EMG records of right and left gluteus medius muscles were recorded to quantify this
synergistic pattern during this fatiguing event. A cross-correlation analysis (see Chapter 2) of left
and right gluteal activity quantified this synergy: a negative correlation resulted when the right
activity increased at the same time as the left activity decreased, and vice versa (for a typical sub-
ject, see Figure 11.5). For this subject, Rxy(τ) had a peak value of −0.677 during a 15-minute block
of the 2-hour trial. The Rxy(τ) is plotted for τ = ±1 s and is fairly flat over that range because of the
ISTUDY
11.3 DYNAMIC BALANCE DURING WALKING 241
τ(s)
0
*Peak Rxy = –0.677 –1 –0.5 0 0.5 1
at T = 0.16 s –0.2
Rxy
–0.4
12
–0.6
10
–0.8 *
8
EMG (% MVC)
6
R glut med
L glut med
4
0
0 10 20 30 40 50 60
Time (s)
Figure 11.5 A 60-seconds linear envelope record of the left and right gluteus medius muscles showing a
synergistic reciprocal pattern. Rxy (upper right) is plotted for τ = ±1 second and showed a peak of −0.677 at
τ = 0.16 second showed these left and right muscles were virtually out of phase, indicating a synergistic load/
unload pattern. See the text for more details on the reported low back pain of those subjects exhibiting this pattern.
fairly long duration of the left and right gluteal bursts of activity. Thus, the peak of Rxy at
τ = 0.16 s reflects what we see from these bursts of activity are virtually 180 out of phase.
Conversely, when the load/unload synergy was not present, there was a positive correlation
[Rxy(τ) = 0.766 during the 15-minute block]; both right and left activity were increasing and
decreasing together, indicating an inefficient co-contraction (see Figure 11.6). Again, the peak
of the Rxy(τ) was at τ = 0.06 s, indicating on this 15-minute record that the left and right gluteal
activity was virtually in phase, and therefore the right and left abductors were fighting each other
to maintain M/L balance.
In this 2-hour workplace study (Nelson-Wong et al., 2008), 23 subjects reported their pain
levels for each 15-minute work task on a visual analog scale, scored from 0 to 40. Fifteen of
the subjects who exhibited this inefficient co-contraction pattern reported a pain score increasing
from an average of 4 in the first work task to 32 in the last task. The remaining eight subjects
exhibited the efficient load/unload pattern and reported an average pain score of 1 in the first task,
increasing slowly to 8 in the last task. There was a significant main effect between these two
groups with a p < 0.0005.
11.3 DYNAMIC BALANCE DURING WALKING
11.3.1 The Human Inverted Pendulum in Steady State Walking

During the gait cycle, there are two periods of single support (each about 40% of the gait cycle)
and two short periods of double-support when both feet are never flat on the ground: the heel
contact foot is about 20 from the horizontal and is plantarflexing rapidly toward a flat foot
position, while the push off heel is well off the ground and weight is entirely on the metatarsals
ISTUDY
0.8 *
*Peak Rxy = 0.766
at T = 0.06 s 0.6
Rxy
0.4
16
0.2
14
0
12 –1 –0.5 0 0.5 1
τ(s)
EMG (% MVC)
10
8
R glut med
6 L glut med
0
0 10 20 30 40 50 60
Time (s)
Figure 11.6 A 60-seconds linear envelope record of the left and right gluteus medius muscles showing an
antagonistic co-contraction pattern. Rxy (upper right) is plotted for τ = ±1 second and shows a peak of 0.766 at
τ = 0.06 second; this shows that these left and right muscles were virtually in phase, indicating an inefficient
coactivation. See the text for more details on the reported low back pain of those subjects exhibiting this pattern.
and toes. When we examine the trajectories of the COP under the feet and the COM of the total
body, we see the challenge to the human control system and more specifically the human as an
inverted pendulum. Figure 11.7 plots the COP and COM of the body during two steps to illustrate
this challenge. The first observation is that the COM never passes within the base of the foot;
rather, it moves forward passing just medial of the inside of each foot. Thus, during each
40% single-support period (LTO to LHC and RTO to RHC) the body is a single-support inverted
pendulum and its horizontal acceleration is decided by the vector joining the COP to the COM
(review the inverted pendulum equations in Chapter 5). Thus, from LTO the M/L trajectory of the
COM is headed toward the right foot, but from the curvature of this trajectory, it is evident that it
is being accelerated medially toward the future position of the left foot. In the plane of progression
(shown by the centerline), there is a deceleration of the COM while it is posterior of the COP and
an acceleration of the COM after the COM moves forward along the foot and is anterior of the
COP. This would be seen in the velocity of the COM, which is decreasing during the first half of
stance and increasing during the latter half of stance. The challenge to the CNS is that the human
is never more than about 400 ms away from falling, and it is the trajectory of the swinging foot
that decides its future position and, therefore, its stability for the next single-support period.
11.3.2 Initiation of Gait

The complete collaboration of both the A/P and M/L muscle groups is never more evident in the
initiation of gait. Going from a very stable balance condition during quiet standing to a walking
state in about two steps requires coordination of the A/P muscles (plantarflexors/dorsiflexors) and
the M/L control (hip abductors/adductors). The goal of these motor patterns during initiation is to
change from the quiet standing pattern to the steady state COP/COM patterns, as described in
Section 11.3.1, in as short a time as possible.
ISTUDY
RHC
RTO
LHC
COM COP
LTO
RHC
Figure 11.7 Trajectories of the COM and COP for two steps during steady state walking. Plotted on the COM
trajectory is shown when key stance events occur: LTO, left toe off; LHC, left heel contact; RTO, right toe off;
RHC, right heel contact. Note that the COM never passes within the base of support of either foot. See the text
for details of the balance challenges to the inverted pendulum model. (Reproduced with permission from
Winter (1991).)
The first major study to track COP/COM profiles during gait initiation yielded the patterns
shown in Figure 11.8 (Jian et al., 1993). Justification for the inverted pendulum model during
this study [Equation (11.3)] yielded correlations that averaged −0.94 in both the A/P and M/L
directions. The origin of both axes is located over the quiet standing position. The initial task
of gait initiation is to accelerate the COM forward and toward the stance limb. Thus, the first
small movement of the COP must be in the opposite direction: posteriorly and toward the swing
limb. This posterior shift is accomplished by a sudden decrease in plantarflexor activity; the lat-
eral shift toward the swing limb results from a momentary loading of the swing limb via increased
activity of the swing limb abductors and decreased activity in the stance limb abductors. As pre-
dicted by Equation (11.3), this momentary shift of the COP accelerates the COM in the desired
direction forward and toward the stance limb. This constitutes the release phase. Interestingly,
this initial increase in swing limb ground reaction force and decrease in stance limb force
appeared in graphs considerably earlier (Herman et al., 1973) but no comment was made. Then
the COP moves rapidly laterally toward the stance limb as the stance limb abductors turn on, the
swing limb abductors turn off, and the swing limb hip flexors/knee extensors accelerate the swing
limb upward and forward. Then, at RTO, the COP is under the left foot and is controlled by the
left leg plantarflexors. The start of this single-support phase is labeled 0%. At this time, the COM
has moved forward about 6 cm, but from its curvature, it is continuing to be accelerated forward
ISTUDY
Top view total body COM versus resultant COP Initiation

–79 walking direction
RHC1
25* 35*
*
0* RTO 20
Z-coordinate (cm)
–5
–69
10+ 20+
–10 25+
0+
–69+ Force plate #2
Quiet standing
–69* COM
–15 COP
–20 Release phase Arrow: CUP→COM
*COP% of gait cycle
Force plate #1 +COM% of gait cycle
27 37 47
X-coordinate (cm)
Figure 11.8 Trajectories of the COM and COP during initiation of gait. Subject was standing quietly with the left
foot on force plate # 2 and the right foot on force plate #1. From the quiet standing position (−69% of the initial gait
cycle), the COP moves posteriorly and laterally toward the swing limb, reaching the release position at −20% of the
gait cycle, then moving rapidly laterally toward the stance foot, reaching it at right toe-off (RTO). The arrows from
the COP to the COM are acceleration vectors showing the acceleration of the COM as predicted by the inverted
pendulum model. See the text for details of the motor control of the COP to achieve the desired COM trajectory.
and now away from the stance foot. During this single-support phase, the stance limb plantar-
flexors increase their activity, causing the COP to move rapidly forward to an initial “toe-
off.” Simultaneously, the right swing limb is swinging forward and RHC1 occurs at 35% of this
first gait cycle. Then, during the double-support phase, the COP moves very rapidly forward
toward the right foot. The trajectory of the COM has now moved forward about 25 cm and
(not shown here) is headed to pass forward along the medial border of the right stance foot
(Jian et al., 1993). Thus, by the end of the first step, the trajectory of the COM has already reached
the steady state walking pattern, as shown previously in Figure 11.7.
The initiation of gait in the young, the elderly, and Parkinson’s patients was analyzed in con-
siderable detail (Halliday et al., 1998). The basic finding in virtually all the kinematic and kinetic
variables was that the temporal patterns were the same but that they were altered by a scaling
factor related to their final steady state walking velocity. The young adults had the highest veloc-
ity; the fit and healthy elderly were slower, and the Parkinson’s patients were slowest. All of the
several significant differences disappeared after the variables were divided by the subjects’ walk-
ing velocity. Thus, we can conclude that the synergistic patterns are still present but are reduced
by some tonic gain control that decreases with aging and disease.
11.3.3 Gait Termination

The challenge to the balance control system during termination of gait is even more critical during
termination of steady state walking. The forward momentum of the body must be removed within
the last two steps, and the COP must be controlled to a position slightly ahead of the COM tra-
jectory as the COM comes to a near stop. Figure 11.9 depicts the COM and COP trajectories
during the contact of the right foot on force plate #2 and then the left foot on force plate #1;
ISTUDY
Top view total body COM versus resultant COP

walking direction Termination
–80
LTO1
LHC Force plate #2

64*
100*
Z-coordinate (cm)
–70
80+ 70+ 64+ 60+

90+ 56+
105* 100+
151* 110+
Quiet standing COM
–60 151+ COP
Arrow: COP→COM
Force plate #1 *COP% of gait cycle
120*
+COM% of gait cycle
34 44 54 64
X-coordinate (cm)
Figure 11.9 Trajectories of the COP and COM during termination of gait. The final two steps are shown; the right
foot on force plate #2 and the left foot on force plate #1. The last gait cycle, from left heel contact (LHC) to LHC, was
0–100%, and shown here is 56–100%. From 56 to 64% is double-support, from 64 to 100% is right single support,
and 100–151% is the final double-support. The COP trajectory moved rapidly forward toward the right foot during
double-support and continues forward under the right foot until LHC, when the left foot begins to load, moving the
COP rapidly laterally until it reaches 120%, when it stops ahead of the COM trajectory. See the text for the
synergistic motor patterns to achieve the desired COM trajectory.
the command to stop was made by a flashing light when the previous left heel made contact with
force plate #3 (not shown). The percent of that stride period began at that heel contact and at 64%
of that stride was LTO1 when 100% of body weight is supported by the right foot. Prior to this
64% point was the double-support period, and the trajectory of the COP moved rapidly forward
from the left foot to the right foot (shown is the 56–64% period). During this double-support
period, the COP moves rapidly forward so that the COP–COM vectors (shown as arrows) indi-
cate a rapid deceleration of COM. Then, during single support of the right foot (64–100%), the
right plantarflexors’ activity increases dramatically, causing COP to move forward to increase the
COP–COM deceleration vector, and by 100%, the COM forward velocity is reduced by about
70%. During this right single-support period, the left swing limb is rapidly decelerated by its hip
extensors/knee flexors, resulting in a step length about half of normal. The rapid loading of the
left foot after LHC causes a rapid lateral and forward shift of the COP until at 120% the COP stops
directly ahead of the trajectory of the COM. This loading of the left foot was achieved simulta-
neously by increased left hip abductor activity and unloading of the right foot by decreased right
hip abductor activity. The final phase, from 120 to 151%, sees the final deceleration of the COM
as the COP moves posteriorly (by reduced plantarflexors of both limbs) and slightly to the right
(by increased right hip abductor and decreased left hip abductor activity) to come to rest at quiet
standing. The most significant aspect of this termination synergy is the proactive control of the
both hip abductors to arrest the later trajectory of the COP at a point directly ahead of the COM
trajectory. About 85% of the forward velocity was arrested by the right plantarflexors and 15% by
the left plantarflexors.
Interestingly, on two unpublished termination trials, one on a patient with peripheral sensory
loss and one on subject with ischemic leg block, the COP trajectory in this final stage when both
ISTUDY
feet were loaded was very erratic: the COP trajectory overshot the COM trajectory in both the
M/L and A/P directions. This indicates that the peripheral sensory system plays a major role
in monitoring and feedback for both the COP and COM positions.
11.4 REFERENCES
Eng, J. J., D. A. Winter, C. D. MacKinnon, and A. E. Patla. “Interaction of Reactive Moments and Centre of
Mass Displacement for Postural Control During Voluntary Arm Movements,” Neurosci. Res. Commun.
11: 73–80, 1992.
Gage, W. H., D. A. Winter, J. S. Frank, and A. L. Adkin. “Kinematic and Kinetic Validity of the Inverted
Pendulum Model in Quiet Standing,” Gait Posture 19: 124–132, 2003.
Halliday, S. E., D. A. Winter, J. S. Frank, A. E. Patla, and F. Prince. “The Initiation of Gait in Young, Elderly
and Parkinson’s Disease Subjects,” Gait Posture 8: 8–14, 1998.
Herman, R., T. Cook, B. Cozzens, and W. Freedman. “Control of Postural Reactions in Man: Initiation of
Gait,” in Control of Posture and Locomotion, R. B. Stein, K. G. Pearson, and J. B. Redford, Eds. (Plenum
Press, New York, 1973), pp. 363–388.
Horak, F. B. and L. M. Nashner. “Central Programming of Postural Movements: Adaptation to Altered Sup-
port Surface Configurations,” J. Neurophysiol. 55: 1369–1381, 1986.
Hunter, I. W. and R. E. Kearney. “Respiratory Components of Human Postural Sway,” Neurosci. Lett. 25:
155–159, 1981.
MacKinnon, C. D. and D. A. Winter. “Control of Whole Body Balance in the Frontal Plane During Human
Walking,” J. Biomech. 26: 633–644, 1993.
Nelson-Wong, E., D. E. Gregory, D. A. Winter, and J. P. Callaghan. “Gluteus Medius Muscle Activation
Patterns as a Predictor of Low Back Pain During Standing,” Clin. Biomech. 23: 545–553, 2008.
Rietdyk, S., A. E. Patla, D. A. Winter, M. G. Ishac, and C. E. Little. “Balance Recovery from Medio-Lateral
Perturbations of the Upper Body During Standing,” J. Biomech. 32: 1149–1158, 1999.
Scholz, J. P. and G. Schöner. “The Uncontrolled Manifold Concept: Identifying Control Variables for a Func-
tional Task,” Exp. Brain Res. 126(3): 289–306, 1999.
Winter, D. A. “Overall Principle of Lower Limb Support During Stance Phase of Gait,” J. Biomech. 13:
923–927, 1980.
Winter, D. A. “Kinematic and Kinetic Patterns in Human Gait; Variability and Compensating Effects,” Hum.
Mov. Sci. 3: 51–76, 1984.
Winter, D. A. The Biomechanics and Motor Control of Human Gait: Normal, Elderly and Pathological,
2nd edition. (Waterloo Biomechanics, Waterloo, ONT, 1991).
Winter, D. A., M. A. Lodge, and W. T. Josenhans. “The Elimination of Respiratory Signals from the Ultralow-
Frequency Ballistocardiogram,” Am. Ht. J. 71: 666–670, 1966.
Winter, D. A., F. Prince, J. S. Frank, C. Powell, and K. F. Zabjek. “A Unified Theory Regarding A/P and M/L
Balance During Quiet Stance,” J. Neurophysiol. 75: 2334–2343, 1996.
Winter, D. A., A. E. Patla, F. Prince, M. Ishak, and K. Gielo-Perzak. “Stiffness Control of Balance in Quiet
Yamagata, M., H. Tateuchi, I. Shimizu, J. Saeki, and N. Ichihashi. “The Relation Between Kinematic Synergy
to Stabilize the Center of Mass During Walking and Future Fall Risks: A 1-year Longitudinal Study,” BMC
Geriatr. 21(1): 240, 2021.
ISTUDY
12
CENTRAL NERVOUS SYSTEM’S ROLE
IN BIOMECHANICS
ALAN R. NEEDLE1,2 AND CHRISTOPHER J. BURCAL3
1
Department of Public Health & Exercise Science, Appalachian State University, Boone, NC, USA
2
Department of Rehabilitation Science, Appalachian State University, Boone, NC, USA
3
School of Health and Kinesiology, University of Nebraska at Omaha, Omaha, NE, USA
12.0 INTRODUCTION
Charles Sherrington’s (1904) work, “Correlation of Reflexes and the Principle of the Common
Path,” established the concept of the motor unit – a lower alpha motor neuron and the muscle
fibers it innervates – as the final common output of the nervous system (Sherrington, 1904).
It is this finding that spurred decades of research whereby electromyography served as the pri-
mary mechanistic measure of the processes of the nervous system during human movement (see
Chapter 9). However, the processes by which the final common output is generated have recently
sparked interest in biomechanists across a variety of contexts. This chapter will focus on these
areas of interest and describe the methodologies used to collect data.
12.1 CENTRAL NERVOUS SYSTEM AND VOLITIONAL CONTROL

OF MOVEMENT
12.1.1 Key Structures for Movement

The sensorimotor system originates at the environmental level (e.g. somatosensory, visual, ves-
tibular afferent information), is integrated within the central nervous system, and then generates
an output along the final common pathway (i.e. motor units). While the anatomy of the central
nervous system is extensive, we can focus on key structures for the production of movement that
offer the potential for measurement or intervention (Figure 12.1). The cerebral cortex, largely
responsible for conscious control of movement offers several important areas. The somatosen-
sory cortex (Brodmann’s area 3, 1, 2), located in the postcentral gyrus is the terminus for sensory
information from somatic receptors. It is located adjacent to the primary motor cortex
(Brodmann’s area 4) in the precentral gyrus, which serves as the origin for motor signals in
the conscious control of movement. The primary motor and somatosensory cortices receive input
from many different cortical areas. Some key cortical areas that interact with the sensorimotor

Companion website:
ISTUDY
248 CENTRAL NERVOUS SYSTEM’S ROLE IN BIOMECHANICS
Primary
Primary somatosensory
Premotor cortex
motor cortex cortex
Supplementary
(BA 4): (BA 3,2,1):
motor cortex
execution sensory
(BA 6)
processing
planning
Basal
ganglia:
tone
coordination
Thalamus:
routing
Brainstem:
patterned and Cerebellum:
automatic comparator
movement
Figure 12.1 Schematic of relevant cortical and subcortical areas for sensorimotor function, including relationships
between these areas. (Created with BioRender.com.)
cortices include the Premotor Area and the Supplementary Motor Area (Brodmann’s area 6) that
will work in a coordinated system with subcortical structures to plan, modify, and execute
motor skills.
It is important to consider that the majority of observed movement happens without significant
conscious control (e.g. gait), with much of this movement being regulated by subcortical areas.
The basal ganglia are a series of nuclei located in cerebral white matter, which include the cau-
date nucleus, putamen, globus pallidus, substantia nigra, and subthalamic nucleus. These nuclei
combine to initiate complex, coordinated movement, largely through the modification of muscle
tone and posture. The cerebellum, while largely associated with vestibular function and control of
balance, offers a key role in movement as the comparator of the nervous system. That is
comparing somatosensory information to an efference copy of the motor action to ensure that
the executed movement matches the planned movement, and offering direct corrections to these
movement patterns. Lastly, the brainstem, consisting of the midbrain, pons, and medulla,
contains a series of direct connections with functional groupings of muscles through descending
tracts capable of controlling posture and muscle tone. The brainstem is the main location for
central pattern generators (CPGs) that coordinate patterned movement (e.g. walking, cycling)
through the rhythmic activation of functional groupings of flexors and extensors. Lastly, the
spinal cord provides more than just the pathway for these signals to reach lower motor neurons
and effector organs. The spinal cord contains dedicated tracts connecting cortical and subcortical
ISTUDY
12.1 CENTRAL NERVOUS SYSTEM AND VOLITIONAL CONTROL OF MOVEMENT 249
structures with both the somatic receptors sensing environmental stimuli and motor neurons to
execute (or inhibit movement). Lastly, the spinal cord gray matter is also the home of a large
number of interneurons capable of integrating sensory stimuli into excitatory or inhibitory
reflexive responses without cortical or subcortical involvement.
12.1.2 Synapses and Neurotransmitters

Neurons are the primary functional unit of the nervous system, with all neurons consisting of a
cell body (soma), axon, and one or more dendrites. Impulses along a neuron travel in a single
direction, with the presynaptic membranes of the dendrites receiving an impulse, transmitting
it to the soma, and then going along an axon to the axon terminal(s). The interaction between
the presynaptic membrane and axon terminals forms the synapse and may be electrical or
chemical in nature. Electrical synapses will transmit voltage across a small gap junction, whereas
chemical synapses rely on the transmission of neurotransmitters or neuromodulators across a
synaptic cleft.
These synapses can be convergent, whereby multiple axon terminals form a synapse at the
same dendrite; or divergent, where the same axon terminal transmits a signal to multiple
dendrites. With regard to motor function, the most relevant neurotransmitters within the central
nervous system include glutamate, gamma-aminobutyric acid (GABA), and glycine. Glutamate is
the primary excitatory neurotransmitter within the central nervous system, whereas GABA is the
primary inhibitory neurotransmitter within the cortex, with glycine being a prevalent inhibitory
neurotransmitter in the spinal cord. Activation of neurons, whether in an electrical or chemical
synapse relies on the generation of an action potential. In the case of chemical synapses, those
neurotransmitters result in an increase or decrease in the postsynaptic voltage, creating excitatory
postsynaptic potentials (EPSPs) or inhibitory postsynaptic potentials (IPSPs) that enable the
propagation of an action potential, constituting a key element of decision making.
12.1.3 CNS Adaptations

A theory posited by Donald Hebb describes the associative learning of neurons that is often
described as, “cells that fire together wire together.” This Hebbian plasticity describes the nature
of synapses to form or increase strength with continued stimuli, or to weaken with the lack of
stimuli (Hebb, 1949). However, there are a number of ways by which synapses adapt that vary
in their speed and durability. Many forms of synaptic facilitation and depression are dependent on
the availability or depletion of calcium ions surrounding the synapse, respectively. The presence
of extracellular calcium enables the transmission of glutamate through higher conductance
N-methyl-d-aspartate (NMDA) channels. This is true for homosynaptic facilitation or depression,
where a previous EPSP either facilitates or diminishes (depresses) a second EPSP, typically
within 200 ms of the initial EPSP. This can further happen over multiple EPSPs, termed post-
tetanic potentiation that can facilitate subsequent potentials for several minutes. Alternately, a
secondary neuron may act on the presynaptic neuron to facilitate or inhibit the release of a neu-
rotransmitter and calcium ions from the presynaptic neuron, thereby strengthening (presynaptic
facilitation, Figure 12.2), or weakening subsequent responses (presynaptic inhibition,
Figure 12.2). The net result of these forms of temporary plasticity are transient changes in the
ability of a neuron to activate – or in the case of motor function – greater (or lesser in the case
of inhibition) muscle activation.
Most forms of synaptic plasticity listed above last from the range of milliseconds to minutes;
however, the goal of most learning is near-permanent changes. This type of learning is facilitated
through long-term potentiation (LTP), or its negative counterpart, long-term depression (LTD).
Our understanding of these processes is best understood as it relates to LTP within the
ISTUDY
(Top)
Facilitatory
neuron
Presynaptic facilitation
Presynaptic Postsynaptic
neuron neuron
(Bottom) Inhibitory
neuron
Presynaptic inhibition
Presynaptic Postsynaptic
neuron neuron
Figure 12.2 (Top) Presynaptic facilitation. A third neuron acts on a synapse, providing input to the presynaptic
neuron that will increase neurotransmitter release and subsequently the likelihood of a subsequent action potential.
(Bottom) Presynaptic inhibition. A third neuron acts on a synapse, providing input to the presynaptic neuron that will
decrease neurotransmitter release and subsequently the likelihood of a subsequent action potential. (Created with
BioRender.com.)
hippocampus; however, the general goal of LTP is to increase the number of NMDA-type glu-
tamate receptors on a postsynaptic neuron (Figure 12.3). Neurons will have AMPA-type gluta-
mate receptors that are permeable to sodium and potassium ions; however, NMDA-type receptors
are permeable to those ions as well as stronger calcium ions. In order for these NMDA-type
receptors to be activated, it requires the removal of a magnesium ion through internal cellular
signaling. Throughout the process of LTP, additional AMPA and NMDA receptors are created
to facilitate action potential transmission from presynaptic to postsynaptic neuron. This process is
the primary mechanism by which new synapses are believed to be formed and strengthened.
12.2 PERIPHERAL NERVOUS SYSTEM AND REFLEXIVE CONTROL

OF MOVEMENT
The role of central nervous system (CNS) structures in generating volitional movement is
described above; however, the peripheral nervous system is also able to independently control
movement through some modulation of reflexive control. While it may be considered peripheral
nervous system control of movement, this is a misnomer as all reflexes require an integrating
center that exists within the CNS (often the spinal cord or brainstem). In fact, all reflexes will
have five components: a sensory receptor, a sensory neuron, an integrating center, a motor neu-
ron, and an effector (e.g. muscle). These differ from volitional control of movement as they are
largely automatic and lack volitional control; however, conscious control can modify reflex
actions, as they are not “hard-wired.” The most basic example of reflexive control comes in
the stretch reflex – a rare monosynaptic reflex. A stretch to
ISTUDY
12.2 PERIPHERAL NERVOUS SYSTEM AND REFLEXIVE CONTROL OF MOVEMENT 251
Glutamate
Ca2+
Mg2+ Na+
New AMPA
receptor
NMDA AMPA
receptor receptor
Internal signaling
Figure 12.3 Long-term potentiation. Consistent firing at a synapse will lead to higher concentrations of calcium
ions and maintain open channels such that internal cell signaling leads to increased NMDA receptors in the
postsynaptic neuron. (Created with BioRender.com.)
(sensory receptor), which transmits its signal along Type Ia/II afferent neurons, is integrated in
the spinal cord with a direct synapse from the sensory neuron to a motor neuron, generates a
response along the α-motor neuron, and causes a contraction of the homonymous muscle. How-
ever, this same stimulus causes multiple responses as the same sensory stimulus gets transmitted
along multiple interneurons within the spinal cord to inhibit the antagonist muscle, constituting a
polysynaptic reflex.
As the stretch reflex is considered the only monosynaptic reflex, it stands to reason that the
remainder of reflexes throughout the nervous system are polysynaptic, with multiple interneur-
ons – those converging and diverging, ascending and descending – generating responses.
While reflexes are trainable and adaptable to various long-term stimuli, there are some common
ones described. For instance, the withdrawal reflex
ISTUDY
glass) that causes contraction of a withdrawal muscle (e.g. hamstring), inhibition of the antag-
onist muscle (e.g. quadriceps), and potentially a stabilizing contraction of the contralateral
extremity.
A number of references above describe control of these reflexes that are not hard-wired.
Among the interneuronal responses to these reflexes include those transmitting descending input
from cortical, cerebellar, and brainstem structures. This descending control will offer the ability
to facilitate or inhibit responses (through presynaptic mechanisms), allowing for reflexes to be
tuned to the specific task. For instance, sitting still and generating a rapid stretch to the Achilles
may facilitate a contraction from the triceps surae; however, if actively stretching the calf, des-
cending control can inhibit the reflexive response and allow the muscle to remain relaxed. These
task-specific adjustments to reflex sensitivity are termed state-dependent reflex reversal,
whereby the same stimulus can generate separate responses based on descending input from
the cortex and other higher structures. This descending control of reflexes forms the basis for
the ability to produce patterned movement (e.g. gait, cycling, swimming) with minimal volitional
input. Within the brainstem and spinal cord are CPG, circuits that when activated pattern move-
ment through reflex modulation. For instance, during gait, a CPG may sense plantar sensation
(stance phase) and recognize that a stretch of the triceps surae should trigger a homonymous
response (push-off ). Meanwhile, with no plantar sensations (swing phase), a similar stretch
of the triceps surae triggers no such response; however, reflexive activation of the tibialis anterior
may occur to facilitate foot clearance.
12.2.1 Sensory Receptors and Motor Units

To understand reflexive control of movement, we must briefly address the types of sensory input
and motor responses in the peripheral nervous system. Rather than a detailed description of all
sensory receptors, we can consider the types of information that are conveyed by these receptors.
We can generally divide the sensory information from receptors into three modalities: proprio-
ception, pressure and touch, and pain. Proprioception consists of the sense of joint position,
movement, and load throughout the body and is largely provided by the muscle spindle (muscle
length, movement), Golgi tendon organ (load), Ruffini endings (ligament stretch), and Pacinian
corpuscles (ligament pressure). These structures will provide some of the most vital information
as it relates to generating and regulating human movement. A number of receptors offer more
crude sensations, including touch and pressure, with these receptors located within the skin
and connective tissue. Examples of these receptors include Merkel disks, Meissner’s corpuscles,
and Pacinian corpuscles. Their specialization comes largely from their ability to discriminate
between fine versus crude touch and texture, and their ability to rapidly adapt or slowly adapt
to changes. Lastly, pain is largely transmitted by free nerve endings throughout the body.
These different modalities should be considered with regard to the measurement of joint sen-
sation. Although the focus of this chapter is techniques to directly measure the CNS, a number
of techniques are used to quantify sensation within biomechanical contexts. Given that propri-
oception includes a sense of joint position, movement, and load; to fully understand proprio-
ception these should each be given consideration. Joint position sense is often quantified
through a subject’s active or passive ability to replicate a given joint angle and quantifying
the error in angle replication (Strong et al., 2021). Movement sense is often quantified as
the threshold-to-detect passive motion, where a device provides a slow movement (~0.1 /sec-
ond), and participants hit a hand switch to indicate they sense movement, with the amount
moved prior to sensing movement used as an outcome measure (Konradsen, 2002). Lastly,
force sense can be determined through force matching tasks, where participants blindly attempt
to replicate a given force level, quantifying the error or variability as an outcome measure
(Wright and Arnold, 2011). For each of the items above, the outcome measure is a form of
“error,” considering the degree to which angle replication is off, the movement offset to detect
passive motion, and the difference between the target load
all of these measurement techniques is that they all
ISTUDY
12.3 METHODOLOGIES TO UNDERSTAND CENTRAL NERVOUS SYSTEM FUNCTION 253
outcome measure (e.g. depression of a hand-switch, actively generating a joint angle, actively
producing a contraction), and therefore does not measure the pure sensory modality of
proprioception (Gandevia et al., 1992, 2006).
Discussions of lower motor neurons as it pertains to electromyographic assessment can be
found in Chapter 9. The focus of that chapter is on α-motor neurons; however, consideration
could be given to γ- or even β-motor neurons. γ-Motor neurons are responsible for the innervation
of intrafusal muscle, from which the muscle spindle is receiving its sensation. It therefore is able
to modify the sensory input to the CNS by increasing or decreasing the sensitivity of the muscle
spindle (Johansson et al., 1991). β-Motor neurons function similarly to γ-motor neurons but
innervate both extrafusal and intrafusal muscle. Electromyographic assessment collects ensemble
activation from all of these motor neurons, with α-motor neurons being the most prominent.
There are very limited in vivo methods to understand the function of these other motor neurons,
although consideration of joint stiffness and muscle tone could be considered. Since γ-motor neu-
rons have a more consistent level of activation, they have the ability to modify muscle tone, or the
baseline tension in muscles. A number of joint stiffness measurement techniques could quantify
this, but caution should be used as the property of muscle tone would be sensitive to the speed of
stretch (Needle et al., 2014).
12.3 METHODOLOGIES TO UNDERSTAND CENTRAL NERVOUS

SYSTEM FUNCTION
Measurement of CNS function in biomechanics research has grown largely over the last several
decades. Understanding the process of sensory integration and movement production has been a
crucial step to advance the scientific understanding of movement and to generate new rehabil-
itation options for a range of musculoskeletal and neurological pathologies. However, there is
often a trade-off between the recording accuracy and the functionality of the measurement tech-
niques. For instance, functional magnetic resonance imaging (fMRI) provides unparalleled spa-
tial resolution to observe CNS function; however, gross joint motion can lead to artifact that is
disruptive to the MRI technology, and temporal accuracy is limited to the length of the scan.
Alternately, electroencephalography (EEG) drastically improves temporal resolution and allows
for some functional movement; however, generally at the cost of spatial resolution. Technolog-
ical innovations are working toward improving the spatial and temporal resolution, especially
during functional tasks. Therefore, consider the following section as a guide for the broad use
of these technologies, in the context of the current evidence.
12.3.1 Functional Magnetic Resonance Imaging (fMRI)

12.3.1.1 Overview of Imaging Techniques. Generally speaking, the gold standard for assess-
ment of cortex activation comes from magnetic resonance imaging technology, in the form of
fMRI. While using magnetic fields to reorient and subsequently measure energy release from
hydrogen ions similar to typical magnetic resonance imaging, fMRI quantifies changes in cere-
bral hemodynamics between resting and functional states. This will quantify the hemodynamic
response function (HRF) using a blood-oxygen-level-dependent (BOLD) contrast. Therefore, the
“cortical activation” described through fMRI represents changes in cerebral blood flow, which is
a correlate of the electrical firing of neurons.
Because of the demands of sugar and oxygen during neuronal firing, a hemodynamic response
allows for an understanding of cortical activation. As oxygenated and deoxygenated hemoglobin
have differing magnetic properties, their response to magnetic field alignment can be detected
using MRI, most commonly using T2∗ weighted acquisitions (Buxton, 2009). Therefore, the
BOLD signal increases as a function of blood flow, blood volume, and blood oxygenation
throughout the cortex in relation to neuronal activity.
in the signal’s acquisition that must be addressed in
ISTUDY
fMRI offers a high level of spatial resolution that allows researchers to identify activation
within specific regions and structures of the cortex. However, this comes with two notable lim-
itations. First, temporal resolution is limited, as full scans last one to several minutes long and
detect the peak HRF seconds after it actually occurs. Therefore, average activation during a task
is explored relative to a control, resting period. Second, in line with the equipment demands, there
are limitations to the level of functional tasks that can be implemented. The head must remain still
throughout the duration of the task, limiting the amount of movement that can be performed, and
any implements used to provide stimuli must be nonmetallic, due to the use of magnetic fields.
12.3.1.2 Experimental Design and Data Collection. While fMRI acquisition will differ
based on the stimulus type and movement paradigm incorporated, there are some standard fea-
tures. As with other physiological techniques, efforts must be made to maximize a signal-to-noise
ratio. In the case of neuroimaging, signal strength can be increased with equipment considera-
tions. The majority of scanners implemented in research to date have utilized a scanner magnet
between 1.5 to 3 T in strength, with scanners up to 7 T currently available. A stronger magnetic
field is able to increase the BOLD signal, though at the cost of possibly increased artifact. Further,
radio frequency (RF) coils, responsible for detecting the energy release, have an impact on image
quality, with more channels yielding higher quality through both spatial and temporal resolution.
Most commonly, 20-, 32-, or 64-channel RF coils are utilized. While these equipment considera-
tions will increase the signal quality, there are numerous considerations to minimize the amount
of noise. While several processing steps will serve this function (described below), stabilization
of the head is generally seen as the key aspect in decreasing extraneous noise, such that signal
quality is improved without changing the MRI hardware itself.
Prior to considering technical configurations and settings of the scan itself, thought must be put
into the task and experimental design. While resting-state fMRI (rs-fMRI) can be collected to
study functional connectivity between cortical regions at rest, most fMRIs used in biomechanical
research are task-related in nature. These tasks could be structured into blocks, in which a task is
performed continuously for 15–30 seconds, with a period of rest between blocks (Figure 12.4a)
(Soares et al., 2016). These blocked approaches represent the most frequently seen design in bio-
mechanical research, as they offer robust power to detect changes. Limitations of this approach
include the ability of a participant to habituate to the task, and a limited ability to link changes to
specific aspects of the task. For instance, in a knee flexion and extension task in which a partic-
ipant straightens and flexes their knee 10 times in a block, it is unclear whether the cortical
activation observed is tied to the knee extension or flexion component. An alternate option is
event-related designs, in which a brief (0.5–8 seconds) stimulus (e.g. sensory pulse, passive
movement) is provided, separated by interstimulus intervals of rest, with analysis attempting
to track the HRF over time (Figure 12.4b). While this increases the temporal resolution of fMRI,
there is a smaller signal-to-noise ratio, and a higher number of trials are needed to adequately
power and detect task-related changes (Soares et al., 2016).
The selection of blocked or event-related designs is largely dependent on the type of task or
stimulus being investigated. Within biomechanics research, there is a challenge to minimize head
movement, while having a participant generate movement and/or providing a sensory stimulus.
There are three common paradigms used to understand cortical activation during movement tasks
in biomechanics research using fMRI. First, simple, isolated movement – often in blocked
designs – can determine cortical activation patterns. This approach requires conscious and careful
stabilization of the head as even ankle movement can be associated with movement artifact at the
head. Thus, a higher number of trials might be needed to ensure enough volumes of high-quality
images are obtained. Second, the use of a projector in combination with mirrors or specialized
goggles can be used to present images or videos to participants depicting movement and
measuring response. Often, these can be used to provide first-person presentation of joint
movement, similar to virtual reality. However, as this approach does not have participants
initiating motion, the signals recorded are largely
ISTUDY
(a) Blocked design
Movement Movement Movement Movement

e.g. repeated flexion/extension
Rest Rest Rest Rest Rest

Time (total = 5 min)
(b) Event-related design

Arrow = Stimulus (e.g. electrical
pulse, passive movement)
Rest
Randomized inter-stimulus interval (ISI)
(0.5–8 s)
Time (total = 30 s)
Figure 12.4 (a) Example of blocked research design. Stimulus is provided or movement is performed
continuously across a block of time and compared to a rest period. (b) Example of an event-related research
design. Stimuli are provided with varying inter-stimulus intervals (ISIs).
(i.e. mirror neurons) and not direct pathways. Additionally, these approaches are largely depend-
ent on ensuring a participant focuses on the images or video presented. A final approach utilized
is motor imagery, where participants are guided to picture themselves producing movement (or
performing a sport-specific task). The limitations of this approach are similar to the use of image
projection, as recordings are dependent on the motor resonance system and require participant
focus, with the additional consideration of standardization of imagery instructions.
With equipment and study design determined, the final considerations are technical configura-
tions and settings of the scan. Gradient- or spin-echo planar imaging using T2∗ weighted MRI
acquisitions is the most common for obtaining whole brain acquisition of cortical activation,
though other options are available (Soares et al., 2016). Each fMRI image consists of more than
100,000 voxels that cover a 2.8–3.5 mm3 area, with a series of 2D images combined to create a
volume (whole brain). Therefore, a number of volumes are collected over time to create a time
series, with the time between volumes referred to as the repetition time (TR). With this under-
standing, we can consider some key decisions, such as the gap between each slice within a vol-
ume, the sequence in which slices may be obtained, and determining an appropriate TR. While
the spatial resolution is improved with a smaller gap between slices, a higher number of slices can
lead to interference and spin history effects that contribute to signal noise. Shorter TRs have the
potential to improve temporal resolution and could be preferred in event-related designs; how-
ever, these shorter times can increase the sensitivity to movement artifact and therefore decrease
signal-to-noise ratio. One final, and key constraint toward
ISTUDY
TABLE 12.1 List of Common Steps Performed During fMRI Analysis

Step Description
Image stabilization Eliminating the first several seconds of a scan to remove inconsistent magnetic
gradients
Slice-timing Corrects for repetition time (TR) across images through interpolation to the first slice or
main image
Motion correction Realignment of each slice to a reference volume, including translations and rotations
Spatial Use of linear and nonlinear methods to align images from individual space to a
transformation referenced standard
Spatial smoothing Spatial smoothing across a 5–10 mm space
Temporal filtering High pass filtering at 0.008–0.01 Hz
First-level analysis Locating voxels that differ between the rest periods and task periods for an individual
subject
Second-level Locating voxels that differ between the rest periods and task periods across a sample
analysis
reflecting the amount of time between the excitation of an ion by magnetic fields and the detection
of that signal, and will typically range from 25 to 40 ms.
12.3.1.3 Processing and Analysis. While a number of sources of artifact can be addressed
through data collection considerations described above, a number of steps must be taken during
preprocessing to optimize the signal-to-noise ratio (Table 12.1). A number of software pipelines
and workflows are available, with the steps described subsequently serving as a guide for what
may be considered, and further reading is recommended to better understand the parameters of
each step (Poldrack et al., 2011; Soares et al., 2016). Image stabilization will often be done by
eliminating the first 10 seconds of any scan, as magnetic gradients may be inconsistent during this
time period. Slice-timing is a technique to normalize the potential offset of TR between images
and will interpolate each image to either the first image or a mean slice. Motion correction will
consist of the realignment of each slice to a reference volume. This may consist of linear steps
including x, y, and z translations and rotations, as well as the process of Scrubbing, where struc-
tured noise can be assessed and removed from the analysis. While translations and rotations are
part of motion correction, these will be further done during a spatial transformation step of fMRI
processing. Spatial transformation will use both linear (translation, rotation, scaling) and nonlin-
ear (warping and distortion maps) methods to align the images from an individual’s space to a
standard reference template. For instance, Talairach space, or the Montreal Neurological Institute
(MNI) template represents standardized models, associated with atlases to identify notable struc-
tures (e.g. automated anatomic labeling, AAL; Harvard-Oxford Atlas). These spatial transforma-
tions are therefore necessary for standardizing across patients and studies and allowing for
detected differences to be associated with a specific cortical region. Finally, a degree of filtering
is warranted, both spatially and temporally. Spatial smoothing might occur across a 5–10 mm
space for group analyses, while temporal filtering might consist of a high pass filter at approx-
imately 0.008 to 0.01 Hz.
Once processed, data must undergo two levels of analysis. The first-level analysis is used to
determine activated areas of the cortex during the task, when compared to resting blocks, within
an individual subject. The second-level analysis consists of compiling data across subjects, in
order to determine the task-related changes that occur across all subjects. Caution must be used
for these analyses, as more than 100,000 voxels of data may be compared, raising the risk of Type
I error regarding findings (Lindquist, 2008). There are several strategies to subsequently decrease
this risk. For instance, many studies may look at activation across regions of interest (ROIs),
which would compile activation of voxels within an
ISTUDY
the number of comparisons. However, this is reliant on a series of a priori ROIs that would be
targeted. Alternately, correction to the error rate may occur, including a family-wise error rate
(FWER), representing a conservative option akin to a Bonferroni correction; or by correcting
the false discovery rate (FDR) that corrects p-values to correct for the risk of false positives,
which is considered less conservative.
A large number of software packages are available that can aid investigators with data proces-
sing and first-level analyses. Analysis of Functional NeuroImages (AFNI) is a tool supported by
the National Institutes of Health that provides tools for smoothing, normalization, and mapping of
data, as well as statistical analyses. The FMRIB Software Library (FSL) is a publicly available
toolkit that provides visualization, normalization, and mapping tools, and is well equipped for
SEM-based analyses. A third, though not final, software package is statistical parametric map-
ping (SPM), a tool that facilitates SPM-based whole-brain analyses of fMRI.
A final step to fMRI analysis includes comparison across groups or time points in a research
study. Though t-tests, analyses of variance, and multiple regression are frequently implemented,
caution is often utilized to minimize the chance of type I error, particularly when performing
voxel-level analyses. At the minimum the level of significance is often lowered to 0.001; how-
ever, this can still lead to 100 voxels providing erroneous data. Therefore, voxel-based thresh-
olding or cluster-extent-based thresholding are often used to lower the rate of false discovery,
with the former considered more conservative (in line with a Bonferroni correction)
(Lindquist, 2008; Soares et al., 2016).
12.3.1.4 Other Imaging Techniques. fMRI represents only one form of advanced imaging
technique for understanding activation of the cortex. While serving as a gold standard in provid-
ing spatial resolution of cortical activation, other MR-based technologies can offer insight into
CNS function and are finding their way into biomechanics research. Diffusion tensor imaging
(DTI) uses MRI technology to track water molecules within the central nervous system over a
period of time (5–20 minutes). These data can then be analyzed with fiber tractography to gen-
erate a structural map of white matter within the brain. The width can provide a basis for con-
nectivity within the cortex, and the atrophy of white matter may be associated with neurological
diseases and even musculoskeletal injury (Soares et al., 2013).
MR spectrography (MRS) determines the chemical composition within a single voxel, allow-
ing investigators to determine the neurometabolites (Rhodes, 2017). MRS is typically a static
scan studying single voxels; however, whole-brain imaging can occur using chemical shift ima-
ging (CSI), allowing for investigation into ROIs in a relatively quick time. MRS can also occur
during functional tasks (fMRS) and allow for tracking of metabolites over time (as fast as ~three
seconds). These imaging techniques have seen limited to no use in biomechanics, but the ability
of MRS, CSI, and fMRS to determine changes in neurotransmitter activity could have notable
benefits for studies implementing motor tasks, and changes throughout learning.
12.3.2 Electroencephalography (EEG)

12.3.2.1 Overview of Electroencephalography. EEG is a method of recording the electrical
potentials generated by populations of neurons from the cerebral cortex. These local potentials
may result from action potentials used for transmission within neurons or postsynaptic potentials,
which reflect the relatively slower process of ion channels opening or closing. Due to the rapid
time scale and propagation of individual action potentials, it is difficult to record this process from
scalp electrodes. Since postsynaptic potentials lead to relatively slower changes in local voltage,
it is possible to record these changes in voltage when large populations of neurons are active and
there is a summation of their individual postsynaptic potentials. It is important to remember that
EEG is recording the electrical activity of tens of billions of cortical neurons which are constantly
communicating locally as well as across regions of the central nervous system. The voltages
recorded from the EEG are small, typically measured in
ISTUDY
Positive or negative fluctuations in voltage that are measured at the scalp electrode represent
activity that: (1) comes from large populations of neurons that are simultaneously active, (2) is
oriented in a similar direction with respect to the measuring electrode, (3) comes from post-
synaptic potentials arising from the same part of the neuron such as the dendrites or cell body,
and (4) the net current flow (e.g. positive or negative) is sufficient enough to overcome opposing
currents (Cohen, 2014; Luck, 2014). These criteria are met by many, but not all, processes that
occur within the brain. Since EEG is recording electrical potentials generated by neurons, it is one
of the few neuroimaging modalities that directly measures cortical activity as opposed to fMRI
which measures change in the local uptake of oxygen.
The summed electrical potentials that makeup EEG combine to produce multidimensional
data. For example, when cortical neurons respond to an external stimulus such as a perturba-
tion, electrical activity may evolve over time. There may be shifts in frequency, where an oscil-
latory brain rhythm becomes more or less dominant in the EEG such as alpha (8–12 Hz) during
periods of alertness or rest. There is also space, as the EEG recorded from each electrode will
differ owing to the proximity of the electrode to the cortical sources and vectors of the electrical
potentials. This phenomenon is best observed when eye blink artifact is present in the EEG as it
will have a greater magnitude in more rostral electrodes closer to the eyes and the magnitude
will decrease in more caudal channels. Information is also contained in the power and phase
components of the overall electrical activity that can be extracted through time-frequency
decomposition. This data can be leveraged to test hypotheses pertaining to dynamic brain activ-
ity. Accompanying this multidimensionality are many analytic approaches that can be applied
to EEG data. It is therefore critically important to have clear scientific objectives when begin-
ning any research project.
12.3.2.2 Experimental Setup. A basic EEG system consists of a recording computer, an

amplifier, electrodes, and an EEG cap which houses the electrodes. The electrodes can be either
active electrodes or passive electrodes, with active electrodes preamplifying the cortical signal
before reaching the digital amplifier and therefore helping to minimize certain artifacts. EEG
caps should be fitted to the participant based on their head circumference to ensure a snug, but
not uncomfortable fit. Most EEG caps have precise electrode spacing according to the inter-
national 10-20, 10-10, or 10-5 system of electrode spacing (Figure 12.5). This allows for con-
sistency across participants during analysis efforts focusing on an individual or array of
electrodes. Efforts should be taken to minimize electromagnetic noise (e.g. cell phones, wire-
less routers) and avoid signal contamination (similar to EMG), which can be accomplished by
collecting data within a Faraday cage. The laboratory climate (temperature and relative humid-
ity) should also be closely controlled to minimize the occurrence of static electricity discharge
which may impart “pops” or square waveforms into the recorded EEG. The air should not be so
dry as to promote the buildup of static electricity and not so warm that it leads to increased scalp
perspiration. Attention should be paid to minimizing any environmental distractions both
within the laboratory as well as the immediate surroundings when possible as these events
can further contaminate the signal. By minimizing environmental sources of noise/data con-
tamination as well as limiting any extraneous sensory, cognitive, or perceptual processes for
the participant, one gains more confidence that the cortical activity being recorded is the result
of experimental conditions and/or manipulations.
One of the most important factors for EEG data analysis is synchronization. Changes in neural
activity without context (e.g. movement performance) lead to speculative conclusions that are not
fully supported by the data. Therefore, it is critical to ensure that the EEG is accurately synchro-
nized with stimuli (auditory, mechanical, visual) or other types of data (EMG, inertial measure-
ment unit (IMUs), kinematics, etc.). Most EEG amplifiers allow for both analog and digital
inputs, enabling researchers to use either custom-made synchronization devices or built-in syn-
chronization output signals from their hardware. Software solutions exist as well, with Lab
Streaming Layer, an open-source software package that
and synchronization across many common hardware
ISTUDY
NASION
NZ
FpZ
Fp1 Fp2
AF7 AF8
AF3 AFZ AF4
F9 F10
F7 F8
F5 F3 F4 F6
F1 Fz F2
FT9 FT7 FT8 FT10
FC5 FC3 FC4 FC6
FC1 FCZ FC2
A1 A2
T9 T7 C5 C3 C1 Cz C2 C4 C6 T8 T10
TP7 CP5 CP3 CP1 CPZ CP2 CP4 CP6 TP8

TP9 TP10
P5 P3 P1 Pz P2 P4 P6
P7 P8
P9 P10
PO3 POZ PO4
PO7 PO8
O1 O2
OZ
IZ
INION
Figure 12.5 The 10-10 system is an international labeling system for describing electrode locations during EEG
research. It allows for standardized electrode locations based on the distance between electrodes. The distance is
measured between the nasion and inion, as well as the distance between the ears at the tragus. Electrodes are
spaced based on 10% of that distance, with the Cz electrode at the midpoint of those two lines. The letters refer
to the lobe of the cortex, and reference channels are often placed on the mastoid process, earlobes, or nose and
are labeled as “A” or “M.” The 10-20 and 10-5 systems may also be adapted for lower and higher density
electrode arrays with 20 and 5% distances between electrodes, respectively.
12.3.2.3 Processing and Managing Electroencephalographic Data. It is very unlikely that

an experiment will be conducted that collects EEG data without artifacts. The nature of EEG
recordings makes them very susceptible to contamination with movement artifacts, line noise
with equipment, perspiration, and many more. Artifacts can even occur from processing unex-
pected environmental stimuli (startle) or natural bodily functions like swallowing saliva. Tak-
ing the time to pilot test helps to identify potential sources of noise and minimize them. While a
number of analysis techniques will aid in the reduction of extraneous noise, the less noise in the
originating signal will lead to ultimately cleaner data, and allow an adequate number of trials to
be maintained. Since many biomechanics experiments inevitably record movement, artifacts
resulting from movement will be one of the largest causes of contaminated or even unstable
data. To ensure the highest signal-to-noise ratio, artifacts in the data need to be addressed,
through filtering, manual or automated artifact rejection, or artifact correction. Each experi-
ment will likely use a unique approach to offline EEG processing; therefore, it is critical that
the exact steps and order of operations are reported clearly when disseminating the results of an
experiment.
Filtering is useful for attenuating artifacts in the data that
Processing the data with a 50 Hz or 60 Hz notch filter can
ISTUDY
(a) (b) (c) (d)

100 μV 100 μV 100 μV 100 μV
F1 FP1 Fz Fz
Fz FP2 F4 F4
F2 AF3 F2 F2
C1 AF4 F6 F6
Cz F3 F8 F8
C2 F4 FT10 FT10
CP1 C3 FT8 FT8
CPz C4 FC6 FC6
CP2 P3 FC4 FC4
P1 P4 FC2 FC2
Figure 12.6 This figure shows three common artifacts that may be encountered when recording EEG. All panels
and EEG channels have a scale of ±50 μV. Panel (a) shows normal EEG clean of artifact. Panel (b) shows an artifact
caused by an eye blink and demonstrates how the amplitude of this artifact decreases as channels move further from
the source. Panel (c) shows EMG artifact caused by neck muscle activity. Panel (d) shows drift that can occur due to
local perspiration under an individual electrode.
high-frequency artifact. High-pass filtering is a useful technique for addressing signal drift that
may occur over time within the amplifier, or because of local skin potentials due to perspiration or
changes in scalp conductance. Band-pass filtering is commonly used in EEG postprocessing prior
to running automated or semi-automated artifact identification, rejection, or correction. Band-
pass filtering also allows for specific bands of activity to be extracted for further analysis such
as event-related (de)synchronization (ERD/S), or spectral power density. Since filters create an
edge artifact due to the transformation applied to attenuate the nondesired frequencies, the begin-
ning, and end of epochs may render the data unusable. Recording an additional 30–60 seconds
before a testing block is recommended to limit this edge artifact.
Filtering struggles to attenuate the impact that transient, stereotyped artifacts have on the EEG
(Figure 12.6). Once common artifacts have been identified (e.g. eyeblinks), the next step is to find
a way to more directly measure this artifact. Many EEG amplifiers allow for simultaneous col-
lection of online digital or analog auxiliary channels. The electrooculogram (EOG) is used to
measure and monitor the potentials created by eye movement and blinks; accelerometers can
be used to monitor perturbations that can create microscopic movements of the EEG electrodes,
and EMG can be used to monitor muscular activity of the head and neck. Most EEG analysis
software packages include several options for manual, semi-automated, and fully automated arti-
fact detection. Once an artifact is identified there are several options. Artifacts can be rejected,
where the segment of data or epoch containing the artifact is removed from the analysis. Artifacts
can be corrected, using one of many different algorithmic approaches that remove the artifact and
attempt to reconstruct the underlying EEG. It is important to remember that each corrected artifact
is a transformation of the original EEG data; the signal to noise ratio is being altered and it is
possible that endogenous EEG activity is being removed any time an artifact is corrected.
A shift toward independent component analysis (ICA)-driven artifact correction has led to less
contamination of the data during artifact correction and results in improved data quality. Pairing
ICA with artifact-specific sensors such as EOG allows for a much more accurate removal of arti-
facts with less risk of removing endogenous brain activity
ISTUDY
TABLE 12.2 Overview of Common EEG Analysis Techniques and their Neurophysiologic
Mechanisms
EEG Technique Measurement Neurophysiologic Mechanism
Event-related Activation from a single or series of Stimulus ascends through spinal tracts
potential electrodes in response to a stimulus to various areas of the cortex
(e.g. postural perturbation, initiation (stimulus specific) and creates a
of movement) patterned response of activity
observed when many trials are
averaged together
Power Spectral EEG is measured continuously through Delta (0–4 Hz), theta (4–7 Hz), alpha
Density a task (e.g. balance, force matching, (8–13 Hz), beta (13–30 Hz), and
etc.). Spectral decomposition is used gamma (30 80 Hz) oscillations are
to determine the spectral power of associated with varying degrees of
specific frequencies at each electrode cortical activation and deactivation.
For instance, alpha frequencies
represent inhibitory thalamocortical
activity, whereas delta frequencies
reflect excitatory cortical activation
Event-related (De) Continuous EEG recordings are Combining the above techniques,
synchronization measured relative to an event (e.g. specific frequency bands are assessed
applied joint load, initiation of relative to a specific event. An
movement, etc.). Time- and phase- increase (ERS) or decrease (ERD) in
locked changes allow investigators to that band’s power relative to a
determine frequency-specific baseline period prior to the event
changes relative to a stimulus determine event-related excitatory or
inhibitory changes
12.3.2.4 Time-Domain Analysis: Event-Related Potentials. One of the most common tech-
niques in EEG research is measuring the event-related potential, or ERP (Table 12.2). This ana-
lytic technique is recommended for testing hypotheses relating to differences in cortical
processing in response to experimentally manipulated stimuli. The high temporal resolution
of EEG can be leveraged to evaluate how cortical activity unfolds as it relates to a time-locked
event. ERPs are often measured using outcomes that reflect the magnitude or timing/latency of a
potential. For example, you may be interested in determining if there is a difference in the mag-
nitude of a motor potential prior to a preplanned movement with varying levels of complexity or
effort and the peak value could be compared between conditions. An averaged ERP also often has
multiple components that are named due to their polarity or latency with respect to the stimulus or
order of components (Figure 12.7).
Processing EEG data to measure an ERP uses the simple analytic approach of averaging
together many trials to increase the signal-to-noise ratio. Since the recorded EEG represents
the sum of all cortical activity, it is very difficult to extract meaningful information from an indi-
vidual trial. The signal (the induced cortical activity as the result of an action or stimulus) is
masked by other ongoing processes. By using a precisely measured temporal event code and
averaging together dozens (sometimes hundreds) of trials, the ERP will be revealed. Therefore,
it is important to ensure that consistent and reliable criteria for creating event markers are used.
Events may be software-driven or hardware-driven, based on input or criteria from other meas-
urement techniques like a force plate or EMG. Since proper synchronization is important in this
method, the accuracy, precision, and reliability of the other measurement modalities should be
taken into consideration when planning an experiment and identifying an appropriate event trig-
ger (Luck, 2014).
ISTUDY
–8
–4 N1
Amplitude (μV)
0
N2
4
P1 P3
8
P2
100 200 300 400

Time after stimulus in milliseconds
Figure 12.7 An example ERP trace reflecting five individual peaks, or components. ERP components are often
named for their directional polarity as well as the order they occur, or the latency at which they are observed. In this
case, P1 could also be named P100. ERPs are conventionally posted with negative values upward.
12.3.2.5 Time-Frequency Analysis Approaches. Decomposing any signal such as EEG into
the time-frequency domain provides researchers with an ability to explore many complex phe-
nomena in the cerebral cortex. For more information on time-frequency decomposition methods
such as Fourier and wavelet transforms, see Chapter 2. The decomposition method is an impor-
tant factor that can significantly impact both the temporal and frequency resolution of your anal-
ysis. In the following section, we will focus on analyses that utilize spectral power. Each analysis
method on its own, much like the decomposition methods, is complex and may have analysis-
specific parameters to be considered (Cohen, 2014).
12.3.2.6 Power Spectral Density. Spectral power provides an overview of the summed mag-
nitude of activity at a given frequency (or range of frequencies) within a window of time. Indi-
vidual frequencies are not often assessed in EEG research, primarily due to differences between
individuals in how their brain processes information. Therefore, a band-based approach is often
used, sometimes called waves (e.g. delta band/waves), wherein a range of frequencies is assessed
in order to determine changes in activity. It is important to clearly define the upper and lower
frequency limits that are being analyzed in the study, as these may differ from research group
to research group. Examining differences in signal power is useful for determining if frequency
band-specific activity is different between conditions of an independent variable. A limitation to
using spectral power to test broader hypotheses in biomechanics comes back to a core principle of
EEG: volume conduction. The EEG signal will contain all the ongoing sensory, motor, percep-
tual, and cognitive processes, therefore clean data that has minimal artifacts is critical. While it
may be possible to mitigate these influences with a high signal-to-noise ratio and subtracting
baseline power from a reference period (see below, event-related synchronization), it is difficult
to draw concrete conclusions using spectral power alone.
12.3.2.7 Event-Related Synchronization. Determining changes in event-related signal power

is similar to the ERP; however, it requires the signal to be decomposed in order to integrate fre-
quency information. This analytic approach has several common names, commonly referred to
as event-related spectral perturbations, or ERD/S. When
ISTUDY
can generate postsynaptic potentials, which can add to the magnitude of activity at a given
frequency. Neurons also communicate information through discharge rate (e.g. frequency),
and thus these two components can combine and reveal information about changes in power
at a given frequency (or band). These changes in power are often used to reflect an increase
or decrease in synchrony among neural populations. Therefore, we can use ERD/S outcomes
to test for changes in cortical activity with respect to an event or stimulus. The practical con-
siderations for calculating ERD/S are very similar to those when designing an experiment that
measures an event-related potential. ERD/S is a referential outcome, meaning it reflects how
change in power unfolds in a window of time across the given frequencies with respect to a
baseline period. Since changes in oscillation reflect a change in a dynamic signal, there may
be a longer refractory period required for the brain to return to its original state. For example,
there is a prolonged change in oscillations that occurs following the completion of a simple dis-
crete movement task; if one were to repeat multiple trials of the same movement over and over, a
short interval between trials would contaminate a referential baseline period with postmovement
neural activity. Since the baseline period is also being decomposed, you should use the same
parameters for both the baseline and period of interest. The signal-to-noise ratio also has impli-
cations for ERD/S analyses; however, since the temporal resolution is relatively lower for time-
frequency analyses than event-related potentials, you often do not need as many trials. Many
studies will report averaging between 50 and 100 trials per condition, and much like conducting
ERP studies, you should a priori set the required number of trials/epochs and report the number
of trials/epochs averaged (Pfurtscheller and Lopes da Silva, 1999).
Careful attention should be given to the part of the experiment that is used for the baseline/
reference period. This is important when calculating ERD/S, or comparing power spectral density
between conditions of a continuous task such as balance. For instance, if an experiment manip-
ulates variables during different gait conditions then it would make more sense to use a baseline
gait condition to capture the normal spectral activity that occurs during gait; this would allow for
differences in cortical activity between conditions to be better evaluated. If using a reference
period from a quiet, seated baseline period prior to the gait conditions, then the change in spectral
activity would reflect both the difference in activity between sitting and gait, as well as any exper-
imental conditions that were manipulated. Therefore, when designing an experiment, it is impor-
tant to factor in comparative conditions and allow for ample time to get a good signal-to-noise
ratio for both the experimental manipulations of interest as well as any baseline/reference periods.
12.3.2.8 Source Localization with Electroencephalography. A difficult question to answer

with EEG is identifying the brain region that is generating the signal being recorded. To answer
this question of source localization, the head is first modeled and then the EEG signal is decom-
posed. Once the head is appropriately modeled, the forward solution is calculated; the forward
solution refers to what the scalp EEG would look like given a known cortical source. The forward
solution is critical for the next step in the processing pipeline when the signal is decomposed to try
to identify the sources generating the potentials recorded in the EEG. This is referred to as the
inverse solution, which aims to identify the most likely cortical source of a signal based on the
EEG. Finding the inverse solution is a computationally intensive process. Many factors can dic-
tate how much each active cortical region contributes to the scalp EEG being recorded, including
current strength, number of neurons active, orientation of the dipole moment, conductivity of
tissues between the source and electrode, and many more that are covered in more detail in a
comprehensive guide by Michel and Brunet (2019).
There are several approaches that can be used to estimate the brain region that is most likely
responsible for the measured scalp EEG. Dipole localization was one of the first methods
described; this method relies on nonlinear optimization equations to identify the best mathemat-
ical location of a single or a small number of dipole moments that are influencing the EEG. This
method is often used with time-locked changes in cortical activity such as ERPs or ERD/S.
A major assumption with this method is that there is only
ISTUDY
a given time, which may not be the case during most biomechanical applications. The other two
main methods are nonadaptive and adaptive distributed source imaging. These methods differ
from the earlier dipole localization methods in that they do not assume a small number of sources,
instead relying on thousands of potential sources in their forward solution model. In both meth-
ods, each location and electrode have a weighting factor that is used to determine the likelihood
that the EEG signal originated from a given source in this 3-dimensional space. Nonadaptive
methods such as low-resolution electromagnetic tomography (LORETA) can offer rapid calcu-
lation of the likely distribution of cortical sources at a given point in time. One drawback to using
nonadaptive methods, such as LORETA, is these methods have relatively lower spatial resolution
compared to other methods, where the data may appear to look like a blurry fMRI (Pascual-
Marqui et al., 1994). This may make it difficult to precisely identify sources if that is the primary
goal of the analysis. Adaptive distributed source imaging methods differ in that these models
assign weights to the EEG data itself in addition to the weights of the electrode positions. Such
methods rely on ICA or principal component analysis (PCA) decomposition to determine the
relative weight that each component contributes to the signal. This additional factor is used
by these models to solve the inverse solution and identify sources. One common method is called
beamforming, which shares the resolution issues of LORETA, but is more sensitive to the
dynamics of the EEG signal due to the integration of signal weights. This added information
could in theory improve the ability to detect components or ERPs of interest that have a smaller
magnitude. A drawback to using adaptive methods such as beamforming is the large number of
parameters that can significantly impact the results of your analysis (Cohen, 2014; Michel and
Brunet, 2019).
Caution should be taken when evaluating results from an investigation using source localiza-
tion due to many of the same factors that affect EEG data collection in the first place. There are
many steps required when conducting these analyses (Michel and Brunet, 2019). While it is the-
oretically possible to achieve high spatial resolution, these would occur under situations where
individual participant-level MRI scans are known, electrode locations are precisely recorded,
head and electrode movement do not occur, the scalp and skull evenly conduct potentials,
and many more. Using a high-density (>32 electrodes) montage is recommended as a minimum
to facilitate the decomposition of the EEG signal, which may improve the ability of your software
to resolve the likely cortical sources and/or dipole locations.
12.3.2.9 Other Analysis Methods for Electroencephalographic Data. There are methods for
EEG analysis that are not covered in this text. Coherence is a common measure that can be used to
determine the relative relationship or consistency between two signals in time and/or phase.
Coherence can be applied to activity occurring at ROI, to look at connectivity, or even to periph-
eral measures to determine the relationship of cortical activity to peripheral efferent signals. Cor-
ticomotor coherence is an example that can show the strength of coupling between motor unit
output/recruitment and cortical rhythms. Coherence can also be used to provide estimates of brain
connectivity. By evaluating the phase-locking properties of signals from different regions and the
evolution of the EEG signal in the time domain, it is possible to use EEG to estimate the dynamic
coordination of brain regions. There are several other methods for determining network connec-
tivity including Granger causality that uses the ability of one signal at predicting the next signal,
mutual information uses linear and nonlinear approaches to detect coupling between regions or
electrodes and is an area of continued algorithmic development and validation. There are dozens
of approaches that can be used when analyzing electrophysiological data, especially with EEG.
Careful thought needs to be applied when determining the best analysis approach to test your
hypothesis, and not fall victim to post hoc over-analysis of your data. It is always best practice
to pilot test both your data collection protocols, as well as, your data analysis procedures and
approach.
ISTUDY
12.3.3 Neural Excitability

12.3.3.1 Overview of Neural Excitability. Neural excitability describes the recruitment and
depolarization properties of motor neurons in response to external stimuli. Whereas EEG and
fMRI are generally a trade-off between spatial and temporal resolution of cortical activation during
a task, measurements of neural excitability describe a largely separate component: the amount of
energy needed to elicit a depolarization of a motor neuron, which is indicative of the resting mem-
brane potentials of that neuron. Further, excitability can also determine the increases in size of
motor responses in relation to the size of the stimulus, often providing a measure of maximal excit-
ability that indicates the total amount of the neuronal pool available. Clinically, neurons that have
higher resting membrane potentials (higher excitability) would be easier to recruit quickly and
possibly effectively during functional tasks. Similarly, neurons that have larger available neuron
pools and better recruitment can lead to larger and stronger muscle contractions during activity.
The counterpart of neural excitability is neural inhibition, or in the case of some of these meas-
urement techniques, intracortical inhibition. Generally speaking, the application of external
energy will activate neurons, leading to a measurable response in the periphery. However, in
the case of the cortex, while stimuli will activate descending excitatory neurons, it will also acti-
vate inhibitory neurons in the cortex. These inhibitory neurons could potentially decrease the
excitability observed through the motor evoked potential (MEP); but alternately can be better
understood through techniques that measure the relative silence following a magnetic pulse,
or the effects of precisely timed conditioning pulses on the MEP. The clinical impact of intra-
cortical inhibition is less clear, although some relationships with levels of resting muscle tone
have been observed.
The implications of neural excitability are not quite as clear as the previous techniques, where
brain activation is recorded during functional tasks. Neural excitability is most often collected at
rest (though not exclusively), and instead informs the researcher about properties of the nervous
system that may translate to movement.
12.3.3.2 Methodologies to Assess Neural Excitability. Assessment of cortical (or corticosp-

inal) excitability is typically performed using transcranial magnetic stimulation (TMS). These
devices utilize dual magnetic fields applied through a double-coil that is placed over the skull.
The magnetic energy from these coils (often up to 2.0 T) is able to painlessly travel through the
scalp and skull and generate a response in a target muscle. There are several types of TMS seen in
the literature, with the focus of this chapter being single-pulse and dual-pulse TMS, often con-
sidered diagnostic TMS. Also observed is repetitive TMS (rTMS) or deep TMS (dTMS), which
differ in the frequency of pulses and coil types and are used for therapeutic purposes rather than
for understanding neural excitability (Groppa et al., 2012).
For both single- and dual-pulse TMS paradigms, protocols will follow a similar procedure.
Outcome measures will be tracked with electromyography electrodes placed over the target
muscle(s). Participants should be familiarized with TMS and the sensations it provides by placing
the coil in the approximate location over the skull, and slowly increasing the intensity of the coil
to the desired intensity, or until a motor response is visible. For investigations into lower extrem-
ity function, the coil should be placed just anterior and lateral to the vertex of the skull (CZ or
C1/C2 of the 10:20 system, See Figure 12.5), while recordings from the upper extremities should
be placed at approximately the C3/C4 or C5/C6 locations contralateral to the target muscle. The
specific hotspot of the target muscle should be located by providing stimuli in a 2.5–5 cm2 radius
while observing recordings of electromyography from the targeted muscle. These responses,
termed MEPs, represent compound muscle action potentials in the target muscle occurring
20–100 ms after the stimulus onset. The hotspot should be identified as the location that provides
the largest MEP at the lowest intensity possible. There are newer methods to identify the hotspot;
however, they require additional technology. For instance, systems can combine motion analysis
ISTUDY
cameras with a headband worn by the participant, and an image of the subject’s brain (from MRI)
or a standard reference patient. While requiring more equipment and expertise, these methods
provide a greater amount of intersession reliability.
Once the hotspot is identified, protocols will vary based on the outcome measure being tar-
geted. Often, investigators will attempt to identify the resting motor threshold (RMT) or active
motor threshold (AMT). While several methods exist to measure this, the difference between
RMT and AMT is whether the participant is relaxed (RMT) as pulses are applied or whether there
is a level of muscle facilitation (AMT), with 10–20% facilitation being common. Common
options for determining motor thresholds include the “5 of 10”’ approach, whereby 10 pulses
are administered at gradually increasing intensities until a visible MEP is observed on 5 of
10 pulses. Alternately, a range of pulses may be applied across a range of intensities, and the
inflection point where motor responses are observed could be inferred as the motor threshold
(Groppa et al., 2012; Hallett, 2007; Needle et al., 2017).
While the RMT and AMT provide information about the depolarization threshold of intracor-
tical neurons, several additional measures can provide additional insight regarding the recruit-
ment of additional neurons. MEP size can be quantified at a given percentage of the RMT or
AMT, with or without muscle facilitation. Further, a stimulus-response curve can be obtained
by providing pulses across a range of intensities ranging from below the RMT to above a maximal
response (Figure 12.8). Multiple pulses should be applied at each intensity, with an increasing
number of pulses leading to a more accurate and reliable curve. The curve of MEP size to stimulus
intensity can then be fitted to a modified Boltzmann equation using a Levenberg–Marquardt algo-
rithm, where y represents the MEP size and x represents stimulus intensity:
MEP max − MEP min

y = MEP min + (12.1)
1 + em I 50 − x
Equation 12.1 above allows for multiple potential dependent measures. While minimum MEP
(MEPmin) represents potential EMG noise or background activation, maximum MEP (MEPmax)
provides a maximum amount of the muscle that can be recruited through cortical activation and
can be normalized/compared to a maximal voluntary contraction or direct nerve or muscle stim-
ulation. Additional parameters include the 50% intensity (I50), indicating the stimulus intensity
where MEP size is 50% of the maximum and the peak slope (m), which represents a parameter
that increases or decreases proportionally to the slope of the curve. The variable e represents
Euler’s number. The I50 and slope together provide valuable information regarding the additional
recruitment of neurons throughout the curve, although the lowest reliability is noted for the slope
parameter. Further single-pulse TMS measures that inform cortical excitability include the central
motor conduction time, representing the conducting delay. This measure should be compared to
the delay from peripheral nerve stimulation, to determine the delay from cortical activation and
no other factors such as electromechanical delay (Devanne et al., 1997).
While TMS is typically not utilized to map brain function, cortical mapping has been done to
determine motor function (Lüdemann-Podubecká and Nowak, 2016). In these protocols, the coil
is systematically moved along a grid while the same sized pulse is provided. The area and inten-
sity of responses across an area can be assessed to understand how brain function changes in
response to a medical condition or intervention. To increase the accuracy of this measure, the
use of technology that incorporates motion analysis with the coil is encouraged, as it allows
for more systematic movements of the coil.
Lastly, single-pulse TMS can be utilized to investigate one measure of intracortical inhibition:
the cortical silent period (CSP). This requires the collection of MEPs during facilitated contrac-
tion (typically 10–20% of maximum effort). As the single-pulse stimulates the excitatory neurons
(generating a MEP), inhibitory neurons are similarly activated, and those mediated by GABAB as
a neurotransmitter can be observed as a period of relative silence following the MEP, lasting
between 150 and 400 ms, termed the CSP (Kimberley et al.,
of more intracortical inhibition that is mediated by GABA
ISTUDY
(a) Stimulus Sample motor evoked potential

0.1
artifact
0.08
0.06
Motor evoked
0.04 potential
Raw EMG amplitude (mV)
(MEP)
0.02
–0.02
–0.04
–0.06
Cortical silent period

–0.08
–0.1
(b) Sample stimulus-response cruve

0.4
0.35 MEPmax
0.3
MEP amplitude (%max)
0.25
0.2 m
0.15
0.1
0.05
MEPmin
I50
0
20 30 40 50 60 70 80 90
Stimulator intensity (% output)
Figure 12.8 (a) Sample motor evoked potential (MEP) during a facilitated trial. (Stirling et al. (2018)/With
permission of Elsevier.) (b) Stimulus-response curve of MEP size versus stimulus intensity. The curve can be
used to extract maximal MEP size (MEPmax), minimal MEP size (MEPmin), slope (m), and the intensity at 50%
of the peak (I50). (Adapted from Needle et al. (2013).)
Dual-pulse TMS differs from single-pulse TMS in that it utilizes dual capacitors, allowing for
customized timing from two pulses provided through the same coil. There are three measures
typically extracted using this technology: short-interval intracortical inhibition (SICI), long-
interval intracortical inhibition (LICI), and intracortical facilitation (ICF). All of these paradigms
utilize a conditioning pulse, followed by a precisely
informing intracortical inhibition from GABAA (SICI),
ISTUDY
(a) Control pulse

Suprathreshold
pulse
(b) Short intracortical inhibition (SICI)

Suprathreshold
pulse
Conditioning
pulse
ISI: 1–8 ms
(c) Long intracortical inhibition (LICI)

Suprathreshold
pulse
Suprathreshold
conditioning
pulse
ISI: 50–200 ms
(d) Intracortical facilitation (ICF)
Suprathreshold
pulse
Conditioning
pulse
ISI: 8–30 ms
Figure 12.9 (a) Example of a normal motor evoked potential (MEP) from a single-pulse of transcranial magnetic
stimulation. (b) Short-interval intracortical inhibition. A sub-threshold conditioning pulse is provided 1–8 ms before
the test stimulus. The relative decrease in MEP size compared to single-pulse is a measure of inhibitory activity.
(c) Long-interval intracortical inhibition. A suprathreshold conditioning pulse is provided 50–200 ms prior to
the test pulse. The decrease in MEP amplitude from the first to second pulse is a measure of inhibitory activity.
(d) Intracortical facilitation. A sub-threshold conditioning pulse is provided 8–30 ms prior to a test pulse.
The increase in MEP amplitude compared to single-pulse is a measure of intracortical facilitation.
NMDA receptors (ICF). In the case of SICI, a subthreshold conditioning pulse (70–80% RMT) is
applied 1–6 ms prior to a suprathreshold test stimulus with the reduction in MEP size describing
the inhibition occurring. In the case of LICI, a suprathreshold conditioning pulse is provided
50–200 ms prior to a suprathreshold test stimulus, with the reduction in MEP size describing
the inhibition occurring. Finally, ICF is quantified by measuring the increase in MEP size with
a subthreshold conditioning pulse applied 8–30 ms prior to a suprathreshold test stimulus
(Figure 12.9) (Chen, 2004; Ridding et al., 1995).
ISTUDY
12.4 PERIPHERAL NERVOUS SYSTEM MEASUREMENT TECHNIQUES 269
TMS is used to understand the connection between the cortex and muscle, and measures are
often described as corticospinal excitability since the stimulus must travel down the corticospinal
tracts to elicit a response at the measurement site. Therefore, it is often done in conjunction with
peripheral electrical stimulation, often in the form of the Hoffmann reflex (H-reflex). The H-
reflex involves stimulation to a peripheral nerve with a similar measurement along a muscle
through electromyography. The nerve should be located by providing supra-threshold pulses
(1 ms width, square wave), and observing where the largest response can be observed at the low-
est stimulation intensity. Pulses should then be applied at that location with gradually increasing
intensities. The electromyographic response from these stimuli can be divided into two cate-
gories: a direct response occurring 5–25 ms after the stimulus (M-wave), and the reflexive
response observed 25–70 ms after the stimulus (H-wave) (Hoffman et al., 2003; Hoffman,
1922). During a typical collection, low-intensity pulses will yield an increase in H-wave ampli-
tude as sensory neurons that transmit muscle spindle activation are larger in diameter and will
trigger a reflex, akin to the stretch reflex. As intensity increases, M-wave amplitude will increase
as the stimulation threshold for motor neurons is reached, while H-reflex amplitude will decrease
due to antidromic inhibition. Stimuli are provided every 10 s until a maximal response is
observed on the M-wave, as indicated by a plateau in the sigmoidal stimulus-response curve
(Figure 12.10). The most common outcome from these measures is the ratio of maximum H-wave
to maximum M-wave (Hmax:Mmax), which is considered a measure of the excitability of the
reflexive motor neuron pool (Hoffman, 1922).
Additional outcome measures can be assessed using H-reflex methodology that could provide
insight into segmental mechanisms. Similar to TMS, investigating the size of motor responses
following the application of a precisely timed conditioning pulse is the key to understanding these
mechanisms. Paired-reflex depression (PRD) uses a conditioning pulse and test stimulus of
15–25% of Mmax intensity, spaced 80 ms apart. Alternately, recurrent inhibition, proportionate
to activation of Renshaw cells within the spinal cord, can be obtained using a conditioning pulse
at 25% of Mmax intensity, and a test stimulus at the Mmax intensity, with pulses provided at both
10 and 80 ms inter-stimulus intervals (ISIs), with the difference between these conditions indi-
cating recurrent inhibition (Earles et al., 2002; Oza et al., 2017).
Collectively, measures of cortical and reflexive excitability and inhibition can offer insight
into the mechanisms that may be contributing to altered motor output, and/or provide mechanistic
data regarding the effects of various interventions on motor function. However, caution should be
used when exploring these components in isolation, as large individual variability may be
observed across populations, and can vary relative to arousal level, recent exercise, or other
extrinsic factors.
12.4 PERIPHERAL NERVOUS SYSTEM MEASUREMENT TECHNIQUES
To fully understand the role of the CNS during movement, one must also consider peripheral
nervous system activity. Much is understood about motor activation within the central nervous
system, as electromyography is the gold standard for understanding alpha motor neuron activa-
tion (See Chapter 9). However, there are a number of additional techniques to understand both
sensory and motor peripheral nervous system function.
12.4.1 Nerve Conduction Studies

As opposed to electromyography that quantifies the amplitude and onset of activation within the
muscle, nerve conduction studies can be used to determine the timing and activation along the
path of a peripheral nerve. As such, these can be investigated both in the context of sensory or
motor stimuli. Similar to electromyography, surface (or sometimes needle) electrodes are placed
along the pathway of a nerve. In the case of a motor study,
ISTUDY
Top
H-wave and M-wave
0.2
0.15
H-wave
0.1
Raw EMG amplitude (mV)
Stimulus artifact
0.05
M-wave
–0.05
–0.1
–0.15
–0.2
Bottom
H-wave and M-wave amplitude over stimulation intensities
1.2
1 Mmax
Peak-to-peak muscle activity (mV)
0.8
0.6
0.4
Hmax
M-wave amplitude
H-wave amplitude
0.2
0
0 10 20 30 40 50 60 70
Intensity (mA)
Figure 12.10 Top: Sample response from peripheral nerve stimulation. The M-wave indicates the direct response
from the muscle as the electrical pulse stimulates the motor nerve. The H-wave indicates the reflexive response as the
afferent nerve is stimulated and initiates a reflexive contraction of the muscle. Bottom: A sample response of H-wave
and M-wave amplitudes as stimulus intensity increases. (Stirling et al. (2018)/With permission of Elsevier.)
brief electrical pulse, similar to that of the H-reflex, while compound motor action potentials
(CMAPs) are collected from sites along the nerve path, quantifying the amplitude and onset
of the response. Stimuli are often applied across a number
to understand where a potential decrease in amplitude or
ISTUDY
12.5 METHODOLOGIES TO UNDERSTAND CENTRAL NERVOUS SYSTEM BEHAVIOR 271
sensory nerve conduction studies rely on stimulation at the distal end of the nerve distribution,
while measuring a similar response, termed the sensory nerve action potential (SNAP) across a
series of sites along the nerve (Mallik and Weir, 2005).
Deficits in CMAP or SNAP amplitude and timing are mostly associated with nervous system
disorders (e.g. spinal cord lesion, multiple sclerosis) but can offer additional insight into the
potential sources of delayed or decreased activation that may be observed through broad
EMG testing, or understanding proprioceptive deficits.
12.4.2 Microneurography
A limitation of understanding sensory function in individuals is that most proprioceptive tests
(e.g. joint position sense, kinesthetic awareness, force matching) rely on motor responses to infer
changes in sensory traffic. One of the few methods to directly understand sensory nerve activity is
through microneurography, consisting of direct recordings of the nerve through a microelectrode
placed directly into the nerve fascicle. This technique is often utilized to understand sympathetic
outflow and its relationship to cardiovascular function but offers several advantages to under-
standing sensory neuron activity (Hagbarth and Burke, 1977).
The microneurographic technique first utilizes ultrasound imaging or electrical stimulation to
locate the approximate location of a nerve. Then, two tungsten microelectrodes are inserted into
the subject – a reference electrode in nearby skin and an active electrode at the nerve location. The
electrodes are connected to a preamplifier and subsequently a nerve traffic analyzer to amplify the
signal. Once inserted, the active electrode will be manipulated by the investigator until it is situ-
ated within the nerve (noted through a discharge signal) and is recorded from the target popula-
tion of neurons. Muscle spindle afferents (i.e. Type Ia, II) are the most common neurons studied
using this technique, although some investigations have explored Golgi tendon organ afferents
(Type Ib) and some have even been performed in Type III/IV afferents. The afferent signal must
then be confirmed to be originating from its source. For muscle spindle afferents, this consists of
(1) responding to passive stretch (2) silence during shortening/muscle twitch, (3) responding to
squeezes of the tendon/muscle belly but not skin brushes, and (4) responding to vibratory stimuli
(Mano et al., 2006). This signal may not only be collected in its raw form, with wavelet decom-
position utilized to extract individual fiber recordings; but also be collected as an ensemble,
where filtering similar to EMG can help to determine the amplitude and timing of responses
(Mano et al., 2006; Needle et al., 2013).
The challenge with the utilization of microneurography in the context of biomechanics
research is the difficulty of coordinating this signal with sensory stimuli and/or movement.
The inserted electrode is often very easy to dislodge, even with the stretch of the skin, and there-
fore only small, controlled motions can typically be implemented. A further methodological
implementation is the potential difficulty in obtaining the desired signal, as the more time spent
searching for the appropriate signal raises the risk of experiencing residual paresthesia following
the protocol.
12.5 METHODOLOGIES TO UNDERSTAND CENTRAL NERVOUS SYSTEM

BEHAVIOR AND ENVIRONMENTAL INTERACTIONS
12.5.1 Virtual Reality

As described in the above sections, many of the techniques for assessing CNS activation in
biomechanics research suffer a trade-off between functional movement and accuracy of record-
ing, as minimizing movement is required to obtain high-quality signals. Virtual reality, often
abbreviated as VR, is a broad term used to reflect computer-generated immersive experiences
that represents a way to improve our ability to understand how the CNS interacts with the envi-
ronment. In research, virtual reality is a strong tool for
ISTUDY
the motor and perceptual consequences of these visuomotor transformations. For example, if
someone stands still while observing a roller coaster in VR, they may unintentionally begin
leaning backward to account for their perceived forward momentum. Due to the manipulation
of optic flow in the above scenario, our brain perceives motion and tries to counteract that
movement. This is the foundation of most VR-focused experiments in biomechanics: lever-
aging the immersive capabilities and temporal resolution of stimulus presentation to observe
changes in motor output.
12.5.1.1 Types of Virtual Reality. VR is most often associated with video games and head-
mounted screens that fully immerse the user in a 3D-rendered scene. Most experiments in VR
will require a custom-coded environment; therefore, it is recommended to consult or include a
software programmer in the research team. There are several types of VR that can be used in
research. Head-mounted VR relies on a headset with precisely shaped lenses that refract the
light from individual computer displays in a manner that creates a three-dimensional projection
of the image that the computer is generating (Figure 12.11). When using a head-mounted VR
headset, the participant will be fully immersed in the scene portrayed on the screen and will not
have any visual cues from their immediate environment to rely on, making them useful if the
goal is to provide a focused visual stimulus. For research purposes, most head-mounted VR
headsets (Figure 12.11) will require a powerful computer capable of rendering the 3D environ-
ment in real-time with minimal latency, often coded using the Unity. One drawback to these
methods is that the individual screens used to generate the 3D environment lead to a limited
vertical and horizontal field of view, which may limit the external validity of the findings
of a study. Normal human vision has 200–220 of horizontal field of view compared to
75–110 in head-mounted VR headsets; vertical field of view is also limited, with normal
vision approximately 130 compared to 90–100 when using a VR headset. It is also possible
to use multiple projectors (at least 2) and large screens to create a virtual environment. This
method has a few advantages to head-mounted VR for biomechanics experiments and is the
most used approach for gait studies that manipulate visual inputs. The common setup for pro-
jection-based VR is a treadmill that is placed a few meters in front of a curved projection screen
or series of flat projection screens (Figure 12.12). Projection-based VR does also limit the field
of view of the participant, but unlike head-mounted VR, the immersive field of view is limited
by the height and width of the projectors and their distance from the participant. Despite the
limitations in how much of the field of view can be rendered virtually, this setup allows the
Figure 12.11 Head-mounted virtual reality can be used to recreate a physical environment. The left panel shows
the physical task that requires moving blocks over a barrier from one side to the other. The middle panel shows the
seated participant interacting with the virtual environment. The right panel shows the participant performing the
same task as the physical task in virtual reality.
ISTUDY
12.5 METHODOLOGIES TO UNDERSTAND CENTRAL NERVOUS SYSTEM BEHAVIOR 273
Figure 12.12 Projection-based virtual reality utilizes a screen, sometimes curved, and one or more projectors to
generate an interactive environment. This method is commonly combined when motion analysis is a primary
measurement of interest to researchers. (Photo provided with permission from Fernando Santos of Bertec LLC.
Laboratory pictured is Bertec Immersive Labs, CoBal Lab at the University of Delaware.)
participant to walk and still perceive cues from their peripheral vision for a more natural gait, as
opposed to head-mounted VR which blocks out all visual input other than the screen. Another
form of virtual reality that can be used for research purposes is augmented reality (AR). The
major difference between AR and VR is that there is no 3D virtual environment, rather the cur-
rent environment the participant is in is supplemented with a digital overlay. Therefore, AR is
not an optimal choice if the goal of the research is to fully manipulate and control the visual
scene. Much like head-mounted VR, a headset, or set of specialized goggles is used for AR, but
the participant can clearly see through the lenses and observe their surroundings.
12.5.1.2 Technical Considerations with VR. There are several important factors to consider
with research that uses VR. From a software side, it is important to ensure that the 3D-rendered
environment is ecologically valid, meaning that 3D-rendered distances between objects and
their digital dimensions are portrayed in a realistic manner. This will help to improve the par-
ticipants’ immersion and perception in the environment so they can focus on the experimental
tasks. This is accomplished by working with skilled software programmers with experience in
the 3D engine and pilot testing iterations. One of the most critical steps in a VR experiment is
familiarizing the participant with VR. Most people have not experienced virtual reality, and the
novelty of interacting with a virtual environment (projection VR) or a 3D scene responsive to
their head movements (head-mounted VR) can significantly impact the focus and attention a
participant is able to afford to the experiment. Familiarization also serves to identify and
potentially limit several other confounding factors that will be outlined in the following sen-
tences. Motion sickness is a common complaint in both head-mounted and projection-based
VR, which can occur acutely and transiently. Due to the potential for transient motion sickness,
frequent breaks should be taken during VR experiments; long-term effects can also occur such
ISTUDY
as migraines. If a participant reports any motion sickness they should be withdrawn from the
study for their safety, and individuals with a history of light sensitivity, vertigo, vestibular con-
ditions (e.g. Meniere’s disease), migraines, or other conditions that may interact poorly with
VR should be excluded from the study. If the experiment requires the participant to perform
a specific task in VR, they should also be familiarized with the task to minimize early learn-
ing/practice effects. It is good practice to report the standardized procedures and tasks that are
used in the experiment, as familiarization should be sufficient to mitigate the potential negative
effects of the novelty of VR environments.
There are also a few hardware and laboratory setup factors that should be considered during
VR research. Since VR is a stimulus presentation modality, experiments will likely be collecting
other forms of data such as kinematics. The headsets that are used for head-mounted VR also
contain accelerometers and IMUs which are used to help track head rotation so the user can inter-
act with the virtual environment. Experiments focused on the neural control of movement may
also be collecting EEG or functional near-infrared spectroscopy (fNIRS) to measure cortical
activity. Since EEG and fNIRS require specialized caps that have precise electrode/optode loca-
tions, you should ensure that the straps that secure the VR headset to the user do not interfere with
the ability to collect high-quality data on CNS function. Cable management is also another factor
that could interfere with data collection, as additional modalities (EEG) or the head-mounted VR
itself often require cables that could alter the way the participant moves. All of these factors can
be addressed through careful planning and in-depth pilot testing prior to beginning an
experiment.
12.6 NERVOUS SYSTEM ROLE IN MUSCLE SYNERGIES
One of the major issues the CNS faces when planning for/executing movement or responding to
perturbations is how to appropriately activate muscles to generate the required torque. Muscle
synergies are a way of describing neuromuscular behavior that attempts to reduce the degrees of
freedom that need to be accounted for during human movement. Synergistic muscle activity
describes a group of muscles working together to complete a movement. Most muscle syner-
gies described in the literature are stereotypical patterns of muscular activity that underly a
behavior, such as gait. One interpretation of muscle synergies is that these patterns are carried
out by subcortical structures, allowing for the cortex to override when movement pattern
flexibility is needed, such as avoiding stepping in a puddle while running or hitting a tennis
forehand while running. Synergies can also be used to understand how the CNS organizes a
reflex-driven response to a support perturbation such as a translation or slip. In this instance,
a group of muscles works together to accomplish a common goal, such as the posterior leg
muscles contracting together to generate a posterior shift of the center of mass following a sup-
port perturbation. The responses to perturbations or threats to stability are used for insight into
how pathology or injury may influence feedback neuromuscular control. Although there is evi-
dence both in support of and opposing the role of muscle synergies, this may be a contentious
topic. You can read more on the role of muscle synergies by Tresch and Jarc (2009). In this
section, we will outline the experimental requirements and basic analysis strategies for muscle
synergy research.
12.6.1 Measurement Techniques and Experimental Setup

The technique of measuring and evaluating muscle synergies relies on EMG. Detailed informa-
tion on how to prepare participants for EMG data collection and appropriate surface electrode
placement can be read in Chapter 9. First, it is important to record data from a large number
ISTUDY
12.6 NERVOUS SYSTEM ROLE IN MUSCLE SYNERGIES 275
of muscles (12 or more), which adds dimensionality to the data. Ideally, the entire limb should be
recorded, as muscles often coordinate and work synergistically across multiple joints. Both ago-
nist and antagonist muscle groups should be recorded as well. The EMG data should be appro-
priately synchronized with all additional sources of data such as perturbation platforms, force
platforms, 3D motion capture systems, accelerometers, or any other equipment that is being used
to measure the performance of the motor task. Perturbation paradigms inherently place the par-
ticipant in an unstable condition, therefore a harness should be used to ensure participant safety.
Pilot testing is important in these experiments and will help you to fine-tune the magnitude and
timing of perturbations. Perturbation research will also require a movable platform, which can
vary greatly in cost; in general, the cost of the platform will rise as the number of controllable
degrees of freedom increases. Muscle synergies are also of interest to researchers during common
movements, especially stereotyped movements such as gait, weightlifting techniques, and reach-
ing tasks. Investigating the coordination of muscular activation can be of interest when research-
ers are concerned with fine motor tasks that involve target manipulation or interception with the
upper and lower extremities.
Experimental design is a critical factor for muscle synergy research. As explained in further
detail by Tresch and Jarc (2009), it is difficult to determine if most synergies reflect neural stra-
tegies for motor control or if they simply reflect the constraints of the task the participants are
performing. One solution to this issue is to test the movement or perturbation of interest under
a broad range of experimental conditions and magnitudes. The ironic solution to this issue can be
accomplished by manipulating the task constraints such as perturbation magnitude or direction,
as well as modifying the speed or force of movement. For example, having a participant bench
press several different loads as a percentage of their 1-repetition maximum would allow for ample
variation in task constraints to assist with identifying the most common underlying muscle syner-
gies. In perturbation studies, steps should also be taken to limit anticipation of support surface
perturbations, including randomizing the order of trials or using a computer to randomly trigger
the perturbation. Due to between- and within-subjects differences in how they may perform a
movement or respond to a stimulus, it is recommended to have participants perform at least
10 trials per movement/stimulus condition as an increased number of trials help to improve
the accuracy of the analysis.
12.6.2 Analysis Techniques

Analyzing EMG data to evaluate muscle synergies does not require many steps. The first two
steps are collecting and preprocessing the EMG data. Researchers should take a similar
approach to setup and data management that are described in Chapter 9. Preprocessing also
includes defining the post-event window (i.e. perturbation, movement onset) of EMG activity
that will be analyzed. The multiple channels of EMG data (i.e. muscles) need to be decomposed
to identify the underlying patterns of activity. There are several approaches that can be used.
The most rudimentary approach is to calculate the onset latency of the given muscles and order
them with respect to a time-locking event such as a perturbation or the gait cycle. More rigorous
methods include ICA/PCA and nonnegative matrix formation. The goal of these analyses is to
determine the underlying patterns of activity that make up the signals being analyzed. The ICA
or PCA assigns a weighting factor to each identified component of activity analyzed through
the signal. This step serves to identify the muscle synergies through mathematical inference.
This approach is recommended as not all synergies have a specific pattern of activation in the
temporal domain, and some muscles may co-activate to generate a specific joint torque rather
than activate in an ordered fashion. The next step is to determine if the EMG data can be
explained using these synergies. The data should also be evaluated for variability both
between- and within subjects, especially when conducting hypothesis-generating experiments
ISTUDY
and observational research. This leads to the calculation of the variance accounted for, often
abbreviated VAF, for which most studies will have an a priori cut-off score set to ensure
the synergy reflects the between- and within-subject variability in the data. It is also possible
to extract the temporal profile of muscle synergies when using nonnegative matrix formation
techniques, which is beneficial during phasic and/or continuous tasks such as gait. A critical
assumption underlying muscle synergy is a strong relationship with variables that represent
movement or performance such as shifts in the center of mass or joint angles. This relationship
should always be confirmed once the candidate synergies have been identified and will assist
when interpreting the findings of the study (Tresch et al., 2009).
12.7 THE CENTRAL NERVOUS SYSTEM AND LEARNING, AND INJURY
12.7.1 Translation of Synaptic Plasticity to Motor Learning

In traditional contexts, motor learning is often inferred from performance, with the retention of
that performance and the transfer of that performance outside of the learned setting serving as the
key elements to infer motor learning has occurred. Further, the timeframe of “learned” adapta-
tions allows for the inference as to whether learning is secondary to long-term (i.e. LTP, LTD) or
short-term changes (e.g. PTP, presynaptic facilitation/depression) in synaptic plasticity.
The implementation of the nervous system measurement techniques described in this chapter
increase the tools for investigators to infer whether learning has occurred. Most typically, as a task
is introduced, it requires a large degree of cortical activation to coordinate and recognize the task.
For instance, learning a new movement pattern requires distinct areas of the motor cortex to activate,
large frontal cortex (e.g. premotor areas, dorsolateral prefrontal cortex) activation to plan and coor-
dinate muscle activation, and perhaps visual cortex activation to visually track the motion. How-
ever, as a task is learned, the amount of activation decreases, as functional groupings may allow for
coordinated muscle activation through less activation, and increased intracortical connections
increase the efficiency of coordinating visual, vestibular, or other relevant control areas (Seidler,
2010). These types of changes are evident both in motor learning studies (tracking individuals
as they learn a task) and in case-control investigations comparing experts in a task to novices.
The central nervous system measuring techniques described in this chapter can offer a number
of ways in which we can gain understanding and confirm these effects. One of the most common
potential decrements is decreased BOLD activation in ROIs through fMRI and/or changes in cor-
tical activation patterns using EEG (e.g. increases in alpha, decreased theta/beta rhythms). While
these represent efficiency in the form of decreased activation, several measurements may reflect
the increased connectivity between areas of the brain. Resting-state fMRI can explore the
activation of various cortical areas without any movement or stimulus, indicating a degree of
functional connectivity between areas that might be responsible for coordinating movement.
Further, DTI and subsequent tractography can describe hypertrophy in intracortical tracts, while
MRS and fMRS can be used to track synaptic strength. Measures of cortical and reflexive
excitability represent some of the most interesting results regarding motor learning – while other
cortical areas are expected to decrease activation – the motor cortex as targeted by TMS may be
likely to increase excitability, represent stronger connections to relevant muscle groups.
12.7.2 Role of Pathology on the Central Nervous System

One final key consideration in biomechanical investigations is the manner in which pathology
affects central nervous system function in the context of movement. For this discussion, we
can divide pathology to include direct lesions to the central nervous system, or upper motor
ISTUDY
12.7 THE CENTRAL NERVOUS SYSTEM AND LEARNING, AND INJURY 277
neuron lesions (e.g. stroke, cerebral palsy, and traumatic brain injury.); as compared to injuries to
the periphery (e.g. musculoskeletal injury). Upper motor neuron lesions have been studied exten-
sively in the context of biomechanics research.
Upper motor neuron lesions represent a constraint (e.g. infarct) directly within the central
nervous system that will typically impede movement in some way, that varies depending on
the area of the cortex affected. Imaging and investigation of the central nervous system will
observe decreased activation and/or excitability within the areas affected. However, given
the plastic nature of the nervous system, activation from surrounding areas may increase
and allow individuals to maintain some function. For instance, following a stroke, an infarct
may decrease the ability of the motor cortex to appropriately activate during gait as well as
decreasing corticospinal excitability; however, increased activation and excitability might
be observed from the contralateral motor cortex. This likely represents strengthening of non-
decussating fibers that activate the impacted muscles, in an attempt to maintain some level of
motor control (Palmer et al., 2016). One final (and important) consideration in the nervous sys-
tem function of individuals with upper motor neuron lesions is the common observations of
spasticity or rigidity. These symptoms are due to a lack of inhibitory input from the central
nervous system in regulating the reflex sensitivity of the muscles surrounding a joint, leading
to a persistent stretch reflex loop. Most biomechanists would observe this symptom through
increased muscle activation and joint stiffness, particularly when observing stiffness relative
to the speed of stretch (Katz and Rymer, 1989). Excitability measures can elucidate these symp-
toms further, as reflexive excitability would notably increase in these individuals, but also dis-
inhibition may be noted within the cortex, as observed through the CSP, SICI, and/or LICI
(Stinear, 2017).
As opposed to pathologies directly impacting the CNS, recent evidence has highlighted a
myriad of changes in the cortex and spinal cord that occur following musculoskeletal injury
(Needle et al., 2017). As opposed to upper motor neuron lesions, there is no constraint within
the CNS itself that limits movement; however, the injury may present a peripheral constraint
that still forces CNS remodeling. For instance, most injuries are associated with an inflamma-
tory response that leads to pain and swelling throughout the joint. These aberrant sensations
decrease the quality of proprioceptive information that reaches the CNS and forces an individ-
ual to coordinate movement with this decreased input (Figure 12.13). Similar negative effects
on proprioceptive feedback can be caused through deafferentation or increased laxity following
ligamentous injury. While these injuries place a sensory constraint on the CNS, pain, and
inflammation also negatively affect motor function through arthrogenic muscle inhibition, evi-
dent through decreased reflexive excitability following musculoskeletal injury, as well as mod-
els that simulate joint effusion (Hopkins and Ingersoll, 2000; Sonnery-Cottet et al., 2019).
Therefore, observations have indicated that an acute injury makes activation of the stabilizing
musculature more difficult due to direct inhibition, and more challenging to coordinate move-
ment from faulty sensory feedback. As such, these factors force the central nervous system to
reorganize in such a way that makes reinjury more likely over the 2–4 weeks following injury.
For instance, with motor excitability decreased, activation of muscle utilizes activation from
extraneous cortical areas (e.g. contralateral motor cortex, premotor areas, visual cortex), allow-
ing for the motion to appear normal, but through atypical processes (Grooms et al., 2016).
However, when outside of a laboratory or clinic, the cortical spread (i.e. increased activation
of extraneous areas with decreased activation of the motor cortex) means that individuals have
fewer cortical resources available for movement flexibility, and become more likely to be
injured when under cognitive load, fatigued, or facing other external constraints (Burcal
et al., 2019).
ISTUDY
Injury event
Peripheral
Pain Inflammation Laxity
deafferentation
Disrupted
sensory
feedback
Continued Induced Altered

flawed neuroplasticity motor
output output
Further
degraded
feedback
Figure 12.13 Paradigm of injury-induced maladaptive neuroplasticity introduced by Needle et al. (2017). Sensory
aberrations secondary to acute and chronic injury perpetuate a cycle of flawed sensory input and motor output that
contributes to subsequent instability and reinjury. (Needle et al. (2017)/With permission of Springer Nature.)
12.8 REFERENCES
Burcal, C. J., A. R. Needle, L. Custer, and A. B. Rosen. “The Effects of Cognitive Loading on Motor Behavior
in Injured Individuals: A Systematic Review,” Sports Med. 49(8): 1233–1253, 2019 https://doi.org/
10.1007/s40279-019-01116-7. PubMed PMID: 31066022.
Buxton, R. B. Introduction to Functional Magnetic Resonance Imaging: Principles and Techniques.
(Cambridge University Press, Cambridge, England, 2009).
Chen, R. “Interactions between inhibitory and excitatory circuits in the human motor cortex,” Exp. Brain Res.
154(1): 1–10, 2004 https://doi.org/10.1007/s00221-003-1684-1. PubMed PMID: 14579004.
Cohen, M. X. Analyzing Neural Time Series Data: Theory and Practice. (MIT Press, 2014).
Devanne, H., B. A. Lavoie, and C. Capaday. “Input-Output Properties and Gain Changes in the Human Cor-
ticospinal Pathway,” Exp. Brain Res. 114(2): 329–338, 1997 https://doi.org/10.1007/PL00005641.
PubMed PMID: 9166922.
Earles, D. R., J. T. Dierking, C. T. Robertson, and D. M. Koceja. “Pre- and Post-Synaptic Control of Moto-
neuron Excitability in Athletes,” Med. Sci. Sports Exerc. 34(11): 1766–1772, 2002 https://doi.org/
10.1249/01.MSS.0000037082.47347.A9. PubMed PMID:
ISTUDY
12.8 REFERENCES 279
Gandevia, S. C., D. I. McCloskey, and D. Burke. “Kinaesthetic Signals and Muscle Contraction,” Trends
Neurosci. 15(2): 62–65, 1992.
Gandevia, S. C., J. L. Smith, M. Crawford, U. Proske, and J. L. Taylor. “Motor Commands Contribute to
Human Position Sense,” J. Physiol. 571(3): 703–710, 2006.
Grooms, D. R., S. J. Page, D. S. Nichols-Larsen, A. M. Chaudhari, S. E. White, and J. A. Onate. “Neuroplas-
ticity Associated with Anterior Cruciate Ligament Reconstruction,” J. Orthop. Sports Phys. Ther.: 1–27,
2016 https://doi.org/10.2519/jospt.2017.7003. PubMed PMID: 27817301.
Groppa, S., A. Oliviero, A. Eisen, A. Quartarone, L. G. Cohen, V. Mall, A. Kaelin-Lang, T. Mima, S. Rossi,
G. W. Thickbroom, P. M. Rossini, U. Ziemann, J. Valls-Sole, and H. R. Siebner. “A Practical Guide to
Diagnostic Transcranial Magnetic Stimulation: Report of an IFCN Committee,” Clin. Neurophysiol.
123(5): 858–882, 2012. Doi: S1388-2457(12)00056-9 [pii] 1016/j.clinph.2012.01.010.
Hagbarth, K. E. and D. Burke. “Microneurography in Man,” Acta Neurol. (Napoli). 32(1): 30–34, 1977.
Hallett, M. “Transcranial Magnetic Stimulation: A Primer,” Neuron 55(2): 187–199, 2007 https://doi.org/
10.1016/j.neuron.2007.06.026. PubMed PMID: 17640522.
Hebb, D. O. The Organization of Behavior. (Wiley, New York, 1949).
Hoffman, P. Untersuchungen über die Eigenreflexe (Sehnenreflexe) menschlicher Muskeln [Studies on the
reflexes (deep tendon reflexes) human muscles]. (Springer, Berlin, 1922).
Hoffman, M. A., R. M. Palmieri, and C. D. Ingersoll. “Simultaneous Hoffmann Reflex Measurements in
Multiple Muscles Around the Ankle,” Int J Neurosci. 113: 39–46, 2003.
Hopkins, J. and C. D. Ingersoll. “Arthrogenic Muscle Inhibition: A Limiting Factor in Joint Rehabilitation,”
J. Sport Rehabil. 9(2): 135–159, 2000.
Johansson, H., P. Sjolander, and P. Sojka. “A Sensory Role for the Cruciate Ligaments,” Clin. Orthop. Relat.
Res. 268: 161–178, 1991. PubMed PMID: 2060205.
Katz, R. T. and W. Z. Rymer. “Spastic Hypertonia: Mechanisms and Measurement,” Arch. Phys. Med. Reha-
bil. 70(2): 144–155, 1989. PubMed PMID: 2644919.
Kimberley, T. J., M. R. Borich, K. D. Prochaska, S. L. Mundfrom, A. E. Perkins, and J. M. Poepping. “Estab-
lishing the Definition and Inter-rater Reliability of Cortical Silent Period Calculation in Subjects with Focal
Hand Dystonia and Healthy Controls,” Neurosci. Lett. 464(2): 84–87, 2009 https://doi.org/10.1016/
j.neulet.2009.08.029. PubMed PMID: 19686807; PMCID: 2752976.
Konradsen, L. “Factors Contributing to Chronic Ankle Instability: Kinesthesia and Joint Position Sense,”
J. Athl. Train. 37(4): 381–385, 2002. PubMed PMID: 12937559; PMCID: 164369.
Lindquist, M. A. “The Statistical Analysis of fMRI Data,” Stat. Sci.: 439–464, 2008.
Luck, S. J. An Introduction to the Event-Related Potential Technique. (MIT Press, Cambridge, MA, 2014).
Lüdemann-Podubecká, J. and D. A. Nowak. “Mapping Cortical Hand Motor Representation Using TMS:
A Method to Assess Brain Plasticity and a Surrogate Marker for Recovery of Function After Stroke?”
Neurosci. Biobehav. Rev. 69: 239–251, 2016 https://doi.org/10.1016/j.neubiorev.2016.07.006. PubMed
PMID: 27435238.
Mallik, A. and A. I. Weir. “Nerve Conduction Studies: Essentials and Pitfalls in Practice,” J. Neurol. Neu-
rosurg. Psychiatry 76(Suppl 2): ii23–ii31, 2005 https://doi.org/10.1136/jnnp.2005.069138. PubMed
PMID: 15961865; PMCID: PMC1765692.
Mano, T., S. Iwase, and S. Toma. “Microneurography as a Tool in Clinical Neurophysiology to Investigate
Peripheral Neural Traffic in Humans,” Clin. Neurophysiol. 117(11): 2357–2384, 2006.
Michel, C. M. and D. Brunet. “EEG Source Imaging: A Practical Review of the Analysis Steps,” Front. Neu-
rol. 10: 325, 2019 https://doi.org/10.3389/fneur.2019.00325. PubMed PMID: 31019487; PMCID:
PMC6458265.
Needle, A. R., C. B. Swanik, W. B. Farquhar, S. J. Thomas, W. C. Rose, and T. W. Kaminski. “Muscle Spin-
dle Traffic in Functionally Unstable Ankles During Ligamentous Stress,” J. Athl. Train. 48(2): 192–202,
2013 https://doi.org/10.4085/1062-6050-48.1.09. PubMed PMID: 23672383; PMCID: PMC3600921.
Needle, A. R., J. Baumeister, T. W. Kaminski, J. S. Higginson, W. B. Farquhar, and C. B. Swanik.
“Neuromechanical Coupling in the Regulation of Muscle Tone and Joint Stiffness,” Scand. J. Med.
Sci. Sports 24(5): 737–748, 2014
ISTUDY
Needle, A. R., A. S. Lepley, and D. R. Grooms. “Central Nervous System Adaptation after Ligamentous
Injury: a Summary of Theories, Evidence, and Clinical Interpretation,” Sports Med. 47(7): 1271–1288,
2017 https://doi.org/10.1007/s40279-016-0666-y. PubMed PMID: 28005191.
Oza, P. D., S. Dudley-Javoroski, and R. K. Shields. “Modulation of H-Reflex Depression with Paired-Pulse
Stimulation in Healthy Active Humans,” Rehabil Res Pract. 2017: 5107097, 2017 https://doi.org/10.1155/
2017/5107097. PubMed PMID: 29225972; PMCID: PMC5684600.
Palmer, J. A., A. R. Needle, R. T. Pohlig, and S. A. Binder-Macleod. “Atypical Cortical Drive During Acti-
vation of the Paretic and Nonparetic Tibialis Anterior is Related to Gait Deficits in Chronic Stroke,” Clin.
Neurophysiol. 127(1): 716–723, 2016 https://doi.org/10.1016/j.clinph.2015.06.013. PubMed PMID:
26142877; PMCID: PMC4684488.
Pascual-Marqui, R. D., C. M. Michel, and D. Lehmann. “Low Resolution Electromagnetic Tomography:
A New Method for Localizing Electrical Activity in the Brain,” Int. J. Psychophysiol. 18(1): 49–65,
1994 https://doi.org/10.1016/0167-8760(84)90014-x. PubMed PMID: 7876038.
Pfurtscheller, G. and F. H. Lopes da Silva. “Event-Related EEG/MEG Synchronization and Desynchroniza-
tion: Basic Principles,” Clin. Neurophysiol. 110(11): 1842–1857, 1999. PubMed PMID: 10576479.
Poldrack, R. A., J. A. Mumford, and T. E. Nichols. Handbook of Functional MRI Data Analysis. (Cambridge
University Press, Cambridge, England, 2011).
Rhodes, C. J. “Magnetic Resonance Spectroscopy,” Sci. Prog. 100(3): 241–292, 2017 https://doi.org/
10.3184/003685017x14993478654307. PubMed PMID: 28779760.
Ridding, M. C., G. Sheean, J. C. Rothwell, R. Inzelberg, and T. Kujirai. “Changes in the Balance Between
Motor Cortical Excitation and Inhibition in Focal, Task Specific Dystonia,” J. Neurol. Neurosurg. Psy-
chiatry 59(5): 493–498, 1995. https://doi.org/10.1136/jnnp.59.5.493.
Seidler, R. D. “Neural Correlates of Motor Learning, Transfer of Learning, and Learning to Learn,” Exerc.
Sport Sci. Rev. 38(1): 3–9, 2010 https://doi.org/10.1097/JES.0b013e3181c5cce7. PubMed PMID:
20016293; PMCID: PMC2796204.
Sherrington, C. S. “Correlation of Reflexes and The Principle of The Common Path,” Report of the British
Association for the Advancement of Science, 74th. Meeting, Cambridge, 1904.
Soares, J. M., P. Marques, V. Alves, and N. Sousa. “A Hitchhiker’s Guide to Diffusion Tensor Imaging,”
Front. Neurosci. 7: 31, 2013 https://doi.org/10.3389/fnins.2013.00031. PubMed PMID: 23486659;
PMCID: PMC3594764.
Soares, J. M., R. Magalhaes, P. S. Moreira, A. Sousa, E. Ganz, A. Sampaio, V. Alves, P. Marques, and N.
Sousa. “A Hitchhiker’s Guide to Functional Magnetic Resonance Imaging,” Front. Neurosci. 10: 515,
2016 https://doi.org/10.3389/fnins.2016.00515. PubMed PMID: 27891073; PMCID: PMC5102908.
Sonnery-Cottet, B., A. Saithna, B. Quelard, M. Daggett, A. Borade, H. Ouanezar, M. Thaunat, and
W. G. Blakeney. “Arthrogenic Muscle Inhibition after ACL Reconstruction: A Scoping Review of the
Efficacy of Interventions,” Br. J. Sports Med. 53(5): 289–298, 2019 https://doi.org/10.1136/
bjsports-2017-098401. PubMed PMID: 30194224; PMCID: PMC6579490.
Stinear, C. M. “Prediction of Motor Recovery after Stroke: Advances in Biomarkers,” Lancet Neurol. 16(10):
826–836, 2017 https://doi.org/10.1016/s1474-4422(17)30283-1. PubMed PMID: 28920888.
Stirling, A. M., M. McBride, E. K. Merritt, and A. R. Needle. “Nervous system excitability and joint stiffness
following short-term dynamic ankle immobilization,” Gait & Posture 59: 46–52, 2018. https://doi.org/
10.1016/j.gaitpost.2017.09.028.
Strong, A., A. Arumugam, E. Tengman, U. Röijezon, and C. K. Häger. “Properties of Knee Joint Position
Sense Tests for Anterior Cruciate Ligament Injury: A Systematic Review and Meta-analysis,” Orthop.
J. Sports Med. 9(6): 23259671211007878, 2021 https://doi.org/10.1177/23259671211007878. PubMed
PMID: 34350298; PMCID: PMC8287371.
Tresch, M. C. and A. Jarc. “The Case for and Against Muscle Synergies,” Curr. Opin. Neurobiol. 19(6): 601–
607, 2009 https://doi.org/10.1016/j.conb.2009.09.002. PubMed PMID: 19828310; PMCID: PMC2818278.
Wright, C. J. and B. L. Arnold. “Review of Eversion Force Sense Characteristics in Individuals with Func-
tional Ankle Instability,” Athl. Train Sport Health Care. 3(1): 33–42, 2011.
ISTUDY
13
A CASE-BASED APPROACH TO
INTERPRETING BIOMECHANICAL DATA
ANKUR PADHYE1, JOHN D. WILLSON2, JOSEPH A. ZENI3, KRISTEN F. NICHOLSON4,
4
AND GARRETT S. BULLOCK
1
Rehabilitation Sciences/Movement Science and Disorders Concentration, Department of Physical Therapy,
East Carolina University, Greenville, NC, USA
2
Department of Physical Therapy, East Carolina University, Greenville, NC, USA
3
Department of Rehabilitation and Movement Sciences, Rutgers University, Newark, NJ, USA
4
Department of Orthopaedic Surgery, Wake Forest School of Medicine, Winston-Salem, NC, USA
13.0 PATELLOFEMORAL PAIN
13.0.1 Introduction
The patellofemoral joint (PFJ) includes the articulation of the patella and the trochlear groove of the
femur. As a sesamoid bone within the quadriceps tendon, the patella increases the ability of the
quadriceps to create knee extension torque by increasing the moment arm of the quadriceps mus-
cles. When the knee is flexed, quadriceps contractions produce retropatellar compressive forces that
can be several times a person’s body weight, particularly during dynamic activities such as running.
These large compressive forces are distributed across the PFJ articular surfaces as a function of knee
and hip joint kinematics in the sagittal, frontal, and transverse plane. Repetitive exposure to a large
compressive force, abrupt changes in training loads, or knee and hip joint kinematics that concen-
trate the distribution of PFJ force to a smaller area (increased PFJ stress) may contribute to patel-
lofemoral pain (PFP), one of the most common musculoskeletal injuries among physically active
populations. Unfortunately, PFP is not a self-limiting condition and symptoms frequently persist for
years following the initial presentation until the proximate cause of the injury is addressed. There-
fore, a thorough clinical and biomechanical evaluation of lower extremity (LE) mechanics during
exacerbating activities is important to design an effective treatment protocol for PFP.
13.0.2 Case Description

The patient was a 35-year-old female with insidious onset of right anterior knee pain 18 months
ago. She was an avid runner over the previous five years, progressively increasing her distance
and intensity, but she was currently unable to tolerate more than a 3 mile run and her current

Companion website:
ISTUDY
282 A CASE-BASED APPROACH TO INTERPRETING BIOMECHANICAL DATA
running volume was 12 miles/week. Her symptoms increased up to 7/10 with running for more
than 10 minutes and increased with repetitive squatting, ascending and descending stairs, and
prolonged sitting (>30 minutes). She denied experiencing instability, locking, numbness, or a his-
tory of trauma to either knee. Her Anterior Knee Pain Scale score was 73% (100% = no symp-
toms). She was given general LE stretching and strengthening exercises and foot orthoses by her
general physician one year ago, but she experienced little benefit. Her motivation to attend phys-
ical therapy included reducing pain so that she can resume regular participation in running and
compete in a half marathon that was six months away.
13.0.3 Patient Examination

The physical examination included LE joint strength and mobility testing. PFJ assessment for
patellar alignment and patellar tracking in open-chain knee flexion/extension did not reveal
any abnormalities. Marked contralateral pelvis drop and medial knee joint displacement were
observed during a single leg step-down activity. The patient also noted apprehension to move-
ment and anxiety during this functional closed chain activity. Such lack of confidence and kine-
siophobia may adversely affect overall rehabilitation potential. In addition, a gait analysis during
running was performed and used both 2D and 3D video systems.
13.0.4 Gait Analysis

The patient was asked to run on a treadmill with separate 2D cameras filming at 120 Hz covering
the anterior–posterior and lateral views and a scaling rod for spatial measurements of her running
mechanics. During running at the patient’s preferred speed (3.2 m/s), she demonstrated a cadence
of 160 steps/minute and an average stance duration of 275 ms. Sagittal plane video analysis
revealed a rearfoot strike pattern well anterior (11 cm) to her center of mass, a large (12 cm) ver-
tical excursion of her center of mass from peak height in swing phase to minimum height in mid-
stance, and only five degrees of knee flexion at initial contact. Frontal plane video during running
revealed increased (eight degrees) hip adduction excursion during the stance phase due to
increased contralateral pelvic drop as well as medial knee displacement of the stance limb, redu-
cing the space between both the knees during mid-stance. Peak PFJ stress and stress impulse dur-
ing running, estimated using 3D motion capture and an inverse-dynamics informed
musculoskeletal model, were 5.2 MPa and 0.059 MPa s, respectively (Figure 13.1). These base-
line PFJ kinetics served as a reference for the comparison of running mechanics following
treatment.
13.0.5 Interpretations and Intervention

The patient demonstrated signs and symptoms consistent with PFJ pain that can be a consequence
of an imbalance of mechanical load (PFJ stress) during running relative to the adaptive/reparative
capabilities of the affected tissues. The intervention developed included a focus on three elements
to reduce PFJ mechanical loading during running and to provide a framework for meeting her
physical activity goals: (1) increase running cadence to reduce step length at her preferred training
velocity (2) address hip and knee joint transverse and frontal plane rotations that reduce PFJ con-
tact area during stance phase, and (3) provide patient education regarding self-management stra-
tegies and load management principles for running exposure progression and regression guided
by symptom response.
To achieve these small alterations to running mechanics that can reduce PFJ mechanical loads
during running, the patient was asked to practice running at her preferred pace while matching the
audio feedback of running cadence using a metronome set at 10% above her preferred cadence
(Bramah et al., 2019). Additionally, to increase awareness of hip and knee joint rotations that
decrease PFJ contact area, a mirror was set in front of the
ISTUDY
13.0 PATELLOFEMORAL PAIN 283
(a) (b)
6 14
12
Hip adduction angle (deg)

5
10
PFJ stress (Mpa)

4
8
3 6
4
2
2
1
0
0 –2
0.00 0.05 0.10 0.15 0.20 0.25
Time (s)
PFJ stress: pre PFJ stress: post
Hip adduction: pre Hip adduction: post
Figure 13.1 (a) Running kinematics at initial contact before (dark) and after (light) adopting increased cadence at
preferred running speed. Note increased knee flexion and heel contact closer to her center of mass. (b) Within-day
changes in patellofemoral joint stress (gray) and hip adduction (black) during the stance phase of running before and
after treatment.
on running kinematics. Step rate feedback and verbal cues such as “increase the space between
your knees” and “keep the waistband of your shorts level” were provided for the first five minutes
of practice to assist the patient successfully increasing step rate and reducing stance phase hip
adduction and internal rotation excursion during running. Over the next five minutes, verbal
and audio cues were removed and provided only as needed. A sample of her running mechanics
while adopting these technique changes was recorded for comparison with her original
presentation (Figure 13.1). Taken together, these adjustments to the running technique reduced
hip adduction excursion by three degrees, decreased peak patellofemoral stress by 19% to
4.2 MPa, and decreased PFJ stress impulse by 25% to 0.044 MPa s during the stance phase.
The patient was educated regarding pain management strategies in the context of a schedule to
progressively increase her running exposure with these recommended kinematic adjustments.
Additionally, she was asked to use a readily available phone application to monitor her compli-
ance with increased cadence while running outside the clinic. Follow-up appointments were
scheduled once a week for two weeks, decreasing to once a month or as needed thereafter.
13.0.6 Patient Outcomes and Discussion

The patient appeared at four total visits over three months. At the end of this time, the patient
reported independence with all aspects of her treatment and a significant reduction in her present-
ing symptoms. 2D video analysis during running at her preferred pace (3.2 m/s) at three months
revealed substantial changes in running mechanics compared to her initial evaluation. In the sag-
ittal view, compared to her initial examination, several hallmarks of a shortened step length were
observed including that step rate increased 10 steps/minutes, stance time decreased to 250 ms,
foot placement was 3 cm closer to her center of mass at initial contact, vertical excursion of
her center of mass decreased 3 cm, and she displayed five degrees more knee flexion at initial
contact while maintaining a rear foot strike pattern. In the frontal plane view, she displayed three
fewer degrees of hip adduction excursion, resulting in decreased medial knee displacement dur-
ing the stance phase. Her Anterior Knee Pain Scale score increased to 95% and she had carefully
increased her running exposure to 24 miles/week.
ISTUDY
The above changes in running kinematics produce kinetic effects culminating in lower per-
step PFJ stress. For example, at a given running velocity, increasing cadence decreases step
length, vertical excursion of the center of mass, and peak vertical ground reaction force. Addi-
tionally, landing with the foot closer to the center of mass reduces posterior ground reaction
force following initial contact. Combined, the reductions in vertical and posterior ground reac-
tion forces lower quadriceps force requirements to generate internal knee extension moment.
Because stance time is also reduced when running with a shorter step length, decreased knee
extension moment is produced over a shorter duration, resulting in a substantially lower angular
impulse and decreased knee joint work (Heiderscheit et al., 2011). These lower quadriceps
requirements during the stance phase reduce mechanical loads at the PFJ in the form of
decreased peak joint contact force and joint force impulse for each step. Further, reducing
hip adduction excursion may reduce iliotibial band tension and laterally directed patellar forces
that can adversely affect the distribution of PFJ contact forces. This improvement in hip adduc-
tion excursion observed after treatment may be a consequence of increased attention and move-
ment feedback training (Willy et al., 2012), increased gluteal muscle contributions during the
late swing and early stance phase with increased step rate (Lenhart et al., 2014), or contribu-
tions from both mechanisms.
13.0.7 Conclusion
The patient in this case demonstrated successful resolution of symptoms and increased tolerance
for running, presumably due to a treatment approach grounded in PFJ mechanics. The case under-
scores the value of a targeted application of biomechanical principles into practice for a person
with PFJ symptoms and clinically identifiable running gait deviations as a proximate cause of her
complaints.
13.0.8 References
Bramah, C., S. J. Preece, N. Gill, and L. Herrington. “A 10% Increase in Step Rate Improves Running
Kinematics and Clinical Outcomes in Runners with Patellofemoral Pain at 4 Weeks and 3 Months,”
Am. J. Sports Med. 47(14): 3406–3413, 2019 https://doi.org/10.1177/0363546519879693.
Heiderscheit, B. C., E. S. Chumanov, M. P. Michalski, C. M. Wille, and M. B. Ryan. “Effects of Step Rate
Manipulation on Joint Mechanics during Running,” Med. Sci. Sports Exerc. 43(2): 296–302, 2011 https://
doi.org/10.1249/MSS.0b013e3181ebedf4.
Lenhart, R., D. Thelen, and B. Heiderscheit. “Hip Muscle Loads during Running at Various Step Rates,”
J. Orthop. Sports Phys. Ther. 44(10): 766–774, A1-4, 2014 https://doi.org/10.2519/jospt.2014.5575.
Willy, R. W., J. P. Scholz, and I. S. Davis. “Mirror Gait Retraining for the Treatment of Patellofemoral Pain in
Female Runners,” Clin. Biomech. (Bristol, Avon) 27(10): 1045–1051, 2012 Dec https://doi.org/10.1016/
j.clinbiomech.2012.07.011.
13.1 BIOMECHANICAL APPROACH TO MANAGE KNEE OSTEOARTHRITIS
13.1.1 Osteoarthritis and Biomechanics

Knee osteoarthritis (OA) is defined by the deterioration of articular cartilage in the tibiofemoral
and/or patellofemoral compartments of the knee joint. Although cartilage loss is the primary
radiographic determinant of OA, this pathology involves many other structures in and around
the knee joint. Notably, there are changes in muscle morphology and function (decreased phys-
iological cross-sectional area, increased intramuscular fat, decreased muscle force production),
sclerosis of subchondral bone, and the abnormal presence of osteophytes (bone spurs)
(Figure 13.2).
ISTUDY
13.1 BIOMECHANICAL APPROACH TO MANAGE KNEE OSTEOARTHRITIS 285
Figure 13.2 Frontal plane radiograph view of the tibiofemoral joint. Note that there is a lack of joint space on the
medial (left) side of the image indicating loss of articular cartilage in the medial compartment. Osteophytes and
subchondral sclerosis are present in the medial compartment. Collectively these findings indicate severe
structural osteoarthritis disease.
Unlike many other pathologies, there is an established link between movement patterns and
the risk of OA initiation and progression. For kinetics, greater adduction moments during gait
have been shown to increase the risk for radiographic OA progression at follow-up (Miyazaki
et al., 2002). As the adduction moment increases, there is a concurrent shift in the distribution
of joint forces in the tibiofemoral interface of the knee; the greater the adduction moment the
greater the load distribution in the medial compartment of the knee joint. Because direct meas-
urement of joint force requires an implanted and instrumented prosthetic device, the adduction
moment is often used as a surrogate measure of internal loads (Kutzner et al., 2013). It is pos-
tulated that this asymmetrical loading creates an environment that fosters cartilage loss, bony
sclerosis, and osteophyte development in the medial compartment; hence OA progression.
From a kinematic perspective, varus and valgus thrusts in the frontal plane have also been
associated with a greater risk of OA progression (Chang et al., 2004). This thrust occurs when
there is a rapid change in a frontal plane joint position toward the knee varus (adduction, bow-
legged position) or knee valgus (abduction, knock-knee position). Rapid changes in frontal plane
position increase the rate of loading in the medial or lateral joint compartments. As these loading
rates exceed the physiological levels, it may also lead to cartilage deterioration and bony changes
within the joint.
Biomechanical changes are not isolated to the frontal plane. Decreased joint excursions in the
sagittal plane during the loading response phase of the gait cycle (knee flexion) have also been
identified as a risk factor for OA progression (Zeni et al., 2019). During the loading phase of the
gait cycle, bodyweight is transitioned from double-limb support to single-limb support. The knee
flexion that occurs immediately after this transition dampens the transition of force on the forward
limb and is thought of as a “shock-absorbing” mechanism during stance. Individuals with knee
OA often present with decreased joint excursion throughout the loading response and maintain
the knee in a flexed position. This may be a consequence or driver of the increased co-contraction
between the quadriceps and hamstring muscles often seen in this population.
Because these biomechanical alterations are part of the disease process or may increase the risk
of OA progression, a biomechanical analysis is an important
ISTUDY
OA. Abnormal kinetics and kinematics can be addressed through the use of movement retraining,
bracing, assistive devices, or other targeted interventions.
13.1.2 Patient History

Case information provided courtesy of Emovi, Inc., Montreal, Canada. The patient in this case
was a 77-year-old male who was diagnosed with “end-stage” or severe medial compartment knee
OA in his left knee. His X-rays revealed a substantial loss of cartilage, subchondral sclerosis, and
osteophytes. Although this patient was physically active and motivated to return to his normal
recreational activities, substantial joint pain prevented him from playing tennis and walking long
distances. This patient followed a typical course of medical management. He received multiple
joint injections for pain relief from his orthopedic physician and participated in several months of
physical therapy. During his therapy, he worked to improve his joint range of motion, muscle
weakness, and swelling in the knee.
After six months of therapy, the patient had improved knee range of motion, small improve-
ments in strength, and a reduction in acute knee pain at rest. However, joint pain during activity
still prevented him from returning to tennis. He continued to report substantial pain during func-
tional weight-bearing activities, including walking for longer than 10 minutes and ascending and
descending stairs. Because of his persistent pain, it was recommended he undergoes a total knee
replacement.
13.1.3 Biomechanical Assessment

In an attempt to delay or prevent the total knee replacement, the patient sought an additional
assessment and opinion on care. To identify movement alterations and potential biomechanical
targets for rehabilitation, the patient underwent a three-dimensional kinematic assessment of the
knee (KneeKG™, Emovi Inc., Canada). As part of this examination, he walked on a treadmill at a
comfortable self-selected pace while knee joint motions in all cardinal planes were evaluated.
This exam revealed several biomechanical risk factors associated with OA progression and
poor function. In the frontal plane, it was found that this individual had a varus thrust, as well
as a static and dynamic varus alignment during standing and walking (Figure 13.3). Peak knee
adduction on the affected limb approached 14 degrees during the early portion of the stance
phase, compared to less than four degrees of adduction in the non-affected knee. The magnitude
of the varus thrust, measured as the difference between the local minimum and maximum values
during the loading response was six degrees. In the sagittal plane, the patient was found to have
limited knee flexion excursion during loading and walked with a characteristic “stiff-leg” gait
pattern (Figure 13.4). This was identified as the reduced arc of motion in the affected knee
(six degrees) compared to the unaffected knee (16 degrees) during the loading response phase
of the gait cycle.
Based on this examination, the therapist developed a targeted home-based exercise program
that focused on neuromuscular retraining to restore more normal movement patterns in the frontal
and sagittal planes. Specifically, these exercises focused on improved control of the knee in the
frontal plane during standing, squatting, and lunging. He was given visual and verbal feedback on
the frontal plane position of the knee as he practiced transitioning weight from the unaffected to
the affected knee. In addition to these exercises, the patient was also given a stabilization brace to
be worn during higher-level activities to limit excessive frontal plane motion and prevent the
varus thrust.
To address the limited sagittal plane joint excursion, the patient was given exercises that
focused on quadriceps strengthening and control in the terminal portion of the range of motion
(0–20 degrees). He performed these exercises on stable surfaces and transitioned to unstable sur-
faces to increase the difficulty of the task over time.
ISTUDY
13.1 BIOMECHANICAL APPROACH TO MANAGE KNEE OSTEOARTHRITIS 287
Frontal plane position (top) and velocity (bottom) of

knee during stance phase of gait
(a) Control limb
OA limb
14
Frontal plane angle (degrees)

12
(+ adduction, – abduction)
10
8
6
4
2
0
–2
0 50 100
(b)
Frontal plane velocity (degrees/percentage)
1.0
(+ adduction, – abduction)
0.5
0.0
–0.5
0 50 100
Percentage of the stance phase
Figure 13.3 Analysis of frontal plane joint angles during walking (a) revealed the patient had considerable knee
varus (adduction positioning) throughout the stance phase of gait on the involved limb. The varus thrust is identified
as the rapid change in joint position during the loading response portion of the gait cycle as the limb transitions from
double-limb support to single-limb support (gray box). The thrust can be seen as the change in position (a) and the
larger adduction velocity (b) in the time-series curves. Note: time in the x-axis has been normalized to 100% of the
stance phase of the gait cycle, so units for velocity are not given in change per second, but rather change per
percentage of stance phase.
The patient returned for a follow-up movement assessment six weeks after his initial evalu-
ation. Results from this examination revealed that the patient was able to walk without a varus
thrust, even without wearing his brace. He also reduced the peak varus alignment during walking
by 40% (8 degrees vs. 14 degrees at baseline) and increased his knee flexion excursion during
loading by nearly 70% (10 degrees vs. 6 degrees at baseline). While these changes are substantial,
the improved movement strategies also led to a reduction in knee pain and function. At a six-
month follow-up, the patient was able to indefinitely postpone his knee replacement and was able
to return to playing tennis with only mild discomfort.
Clinicians should consider a biomechanical assessment when working with patients with
knee OA. Progression of knee OA is driven, in part, by biomechanical markers. Selecting
ISTUDY
Knee flexion angle during stance phase of gait

60 Control limb
OA limb
Knee flexion angle (degrees) 40
20
0
0 50 100
Percentage of the stance phase
Figure 13.4 Analysis of sagittal plane joint angles during walking revealed the patient had considerably less knee
flexion excursion during the loading phase and maintained the knee in a flexed position throughout the stance phase
of gait.
exercises and interventions to address abnormal mechanics may lead to better long-term out-
comes and preserve the structural integrity of the joint.
13.1.4 References
Chang, A., K. Hayes, D. Dunlop et al. “Thrust during Ambulation and the Progression of Knee
Osteoarthritis,” Arthritis Rheum. 50(12): 3897–3903, 2004.
Kutzner, I., A. Trepczynski, M. O. Heller, and G. Bergmann. “Knee Adduction Moment and Medial Contact
Force--Facts About Their Correlation during Gait,” PLoS One 8(12): e81036, 2013.
Miyazaki, T., M. Wada, H. Kawahara, M. Sato, H. Baba, and S. Shimada. “Dynamic Load at Baseline Can
Predict Radiographic Disease Progression in Medial Compartment Knee Osteoarthritis,” Ann. Rheum. Dis.
61(7): 617–622, 2002.
Zeni, J. A., Jr., P. Flowers, M. Bade, V. Cheuy, J. Stevens-Lapsley, and L. Snyder-Mackler. “Stiff Knee Gait
May Increase Risk of Second Total Knee Arthroplasty,” J. Orthop. Res. 37(2): 397–402, 2019.
13.2 ULNAR COLLATERAL LIGAMENT RECONSTRUCTION
Overuse conditions and throw-related injuries are a significant problem at the shoulder and elbow
for athletes, particularly baseball pitchers (Conte et al., 2016; Hootman et al., 2007; Posner et al.,
2011; Shanley et al., 2011). Elbow injuries are often attributed to excessive elbow valgus torque
(Pytiak et al., 2018; Agresta et al., 2019; Anz et al., 2010), which can reach 115 N m when
pitching. The ulnar collateral ligament (UCL) is designed to reduce valgus motion at the elbow
and can become injured or torn with repetitive pitching activities (Feltner and Dapena, 2016;
Werner et al., 1993, 2002). Modern biomechanical pitching analysis can identify risk factors
associated with elbow injury. Clinicians and researchers can use this information to reduce injury
and improve performance through targeted training activities.
Three-dimensional motion capture has been used to examine how full-body kinematics influ-
ence joint stresses and pitching velocity (Agresta et al.,
ISTUDY
13.2 ULNAR COLLATERAL LIGAMENT RECONSTRUCTION 289
et al., 2020). While there are hundreds of variables of interest when analyzing pitching, pitch
velocity was the most influential variable, and certain kinematic variables revealed significant
influence on elbow stress (Nicholson et al., 2021). Some of these variables include maximum
shoulder external rotation, ground reaction force, shoulder abduction, and humeral rotation
velocity. Clinicians and coaches can focus on these variables and provide feedback to change
pitching mechanics and, ultimately, reduce elbow stress. Decreasing shoulder abduction at foot
strike (Werner et al., 2002; Aguinaldo and Escamilla, 2019) and increasing shoulder external
rotation torque (Aguinaldo et al., 2007; Sabick et al., 2004; Sgroi et al., 2015) (achievable by
increasing maximum external rotation and/or external rotation velocity) have been shown to
reduce elbow valgus load when pitching.
Evaluating and interpreting key biomechanical variables is an important part of pitching anal-
ysis. The purpose of this case is to explore an approach to interpreting biomechanical pitching
data to limit elbow stress. While every individual is unique, this case demonstrates some of the
more common deviations and variables of interest that can be identified and corrected through
biomechanical analysis.
13.2.1 Player History

A 22-year-old right-handed relief pitcher in a Division I college baseball program presented for a
3D biomechanical pitching assessment. The pitcher had suffered a season-ending injury to the
UCL of his elbow in 2018. He underwent UCL reconstruction and missed the 2019 season
but was cleared by his sports medicine physician to return to throwing. The 3D biomechanical
pitching assessment was conducted in August 2019, about 14 months after surgery. Two weeks
after the assessment, the pitcher re-tore his UCL during competitive play.
Elbow valgus torque is a measure of the torque that is created in resistance to elbow valgus
motion and is considered an indicator of stress in the UCL. As discussed in previous chapters,
joint torques are influenced by height and body weight. To account for these anthropometric
variables, elbow valgus torque is normalized by body weight times height (BW×H) during
the analysis. In our experience, elbow valgus torque that exceeds 5% BW×H should be reduced
to decrease UCL stress. This pitcher presented with maximum elbow valgus torques between
6 and 8% BW×H when throwing.
A cursory examination of the key performance and injury indicators at specific points in the
pitching cycle (Table 13.1) indicates that the pitcher had excessive shoulder external rotation
motion, while he had lower than normal values for hip–shoulder separation at footstrike, knee exten-
sion from footstrike to ball release, and shoulder internal rotation angular velocity. While some of
these metrics do not directly indicate an increase in elbow stress, they warrant further exploration.
When evaluating key performance and injury indicators, a bottom-up approach evaluates these
metrics from the legs first, followed by the trunk and upper extremities. When landing on the lead
leg, this pitcher landed in a “closed” position where the landing leg crossed the midline. Ideally,
the lead foot should land in line with the drive (rear) leg while also being slightly internally
rotated (Dillman et al., 1993). Landing in the closed position reduces throwing power and places
the pitcher with an unstable base of support (Figure 13.5).
There were also concerning findings in sagittal plane knee motion in the lead leg. It is recom-
mended that the lead leg moves through at least 10 degrees of knee extension excursion from
footstrike to ball release (Figure 13.6). Extending the knee at this point in the delivery allows
a pitcher to use the powerful quadriceps and gluteal muscles before ball release to create a catapult
effect with the knee and hip. This increases ball velocity without increasing elbow valgus torque.
In this case study, the pitcher continues to flex his knee from footstrike to ball release. This may
be the result of altered motor control patterns or impairments with muscular strength. Knee flex-
ion during this period reduces the lower extremity’s contribution to velocity, which requires the
pitcher to generate more energy from the upper extremity to maintain ball velocity. It is possible
that this contributes to excessive torques at the elbow.
ISTUDY
TABLE 13.1 Key Performance and Injury Indicators at Maximums, Footstrike, or Ball Release
Off
Breaking Ball Fastball Speed Reference
(n = 3) (n = 3) (n = 1) (n = 261)
Ball velocity 78.2 ± 0.3 mph 83.8 ± 0.7 mph 80.0 mph 76.0–85.3 mph
Stride length 74.9% H ± 0.4 H 71.2% H ± 0.5% H 72.8% H 69.5–86.7% H
Shoulder abduction at footstrike 94.4 ± 0.4 96.0 ± 0.2 96.9 76.5 –100.2
Shoulder external rotation 194.8 ± 1.6 194.2 ± 1.9 192.5 156.7 –179.3
maximum
Hip–shoulder separation at 43.0 ± 1.1 40.3 ± 1.2 40.6 38.3 –55.3
footstrike
Shoulder horizontal abduction −68.9 ± 0.5 −67.9 ± 0.5 −68.8 −35.9 to −16.1
at footstrike
Elbow flexion at footstrike 94.3 ± 3.6 93.1 ± 1.8 94.2 73.2 –104.3
Elbow flexion at ball release 47.3 ± 1.7 44.4 ± 5.2 60.8 20.1 –36.6
Lead knee flexion at footstrike 49.7 ± 0.8 48.7 ± 2.2 49.3 38.4 –57.4
Lead knee flexion at ball release 43.0 ± 2.8 48.0 ± 2.2 51.0 28.4 –59.8
Pelvis rotation angular velocity 650.0 ± 6.3 /s 652.4 ± 5.1 /s 661.8 /s 596.9–788.9 /s
Trunk rotation angular velocity 1156.0 ± 5.6 /s 1162.3 ± 5.2 /s 1166.5 /s 995.7–1174.3 /s
Shoulder horizontal adduction 1009.8 ± 16.2 /s 952.4 ± 28.0 /s 923.9 /s 218.3–1090.3 /s
angular velocity
Shoulder internal rotation 4604.7 ± 196.6 /s 4371.3 ± 174.4 /s 4589.8 /s 4638.7–5751.3 /s
angular velocity
Lead leg GRF maximum 2.1% W ± 0.04% W 2.1% W ± 0.01% W 2.1% W 1.9–2.5% W
Drive leg GRF maximum 1.2% W ± 0.2% W 1.3% W ± 0.2% W 1.3% W 1.1–1.5% W
References are lab-generated norms for collegiate baseball pitchers reported as a range of mean ± one standard deviation.
Breaking balls are curveballs or sliders while off-speed pitches are changeups. Mean ± standard deviation. H, height;
W, weight.
Figure 13.5 The pitcher’s stride leg crosses over midline with the shank angled toward first base. This is not a
powerful position, and the pitcher lacks a stable base for rotation and power generation.
ISTUDY
150°
100°
Flex +, Ext –
50°
0°
0s 0.2s 0.4s 0.6s 0.8s 1s 1.2s 1.4s

Setup to release + 100 ms
Figure 13.6 Knee flexion graph. The first vertical lines indicate footstrike. The second vertical line indicates ball
release. It is recommended to have at least 10 degrees of knee extension from footstrike to ball release. Gray band
represents ± one standard deviation of the reference normal. This graph represents the knee action of the pitcher in
this case example. He is landing with about 50 degrees of knee flexion/bend. He flexes his knee further before
achieving some extension, but by ball release, he is in a more flexed position (51 degrees) than at footstrike.
(a) (b)
50°
Ccw –, Cw +
25°
0°
–25°
0s 0.2s 0.4s 0.6s 0.8s 1s 1.2s 1.4s

Figure 13.7 (a) Moment of maximum hip–shoulder separation. Pitchers who exhibit maximum hip–shoulder
separation angles of greater than 55 degrees have pelvis facing home at maximum hip–shoulder separation. (b)
Hip–shoulder separation graph. The first vertical lines indicate footstrike. The second vertical line indicates ball
release. Gray band represents ± one standard deviation of the reference normal. This pitcher’s maximum hip–
shoulder separation is less than 55 degrees (arrow).
Hip–shoulder separation is another important metric related to elbow torque and is measured
as the angle in the transverse plane created between the frontal plane alignment of the pelvis and
frontal plane alignment of the shoulders. This angle is a key indicator of lower extremity loading;
greater angles create a larger spring effect between the lower extremity and trunk. When eval-
uating throwing mechanics, a pitcher should have at least 55 degrees of hip–shoulder separation
just after stride leg footstrike. The pitcher in this case had a maximum hip–shoulder separation of
45 degrees (Figure 13.7). Decreased hip–shoulder separation limits trunk rotation velocity and
potentially pitch velocity (Bullock et al., 2020a). To maintain appropriate pitch velocity, a pitcher
with limited hip–shoulder separation may use other centripetal force generation mechanisms at
the shoulder and elbow, which can increase detrimental elbow torques (Hirashima et al., 2002).
For this pitcher, he compensated with excessive shoulder horizontal abduction (Figure 13.8).
This is described as “throwing out of the scapular plane” and can contribute to shoulder and
elbow injuries.
ISTUDY
(a) (b)
100°
50°
Abd –, Add +
0°
–50°
–100°
0s 0.2s 0.4s 0.6s 0.8s 1s 1.2s 1.4s
Figure 13.8 (a) Moment of maximum shoulder horizontal abduction. (b) Shoulder horizontal abduction graph.
The first vertical lines indicate footstrike. The second vertical line indicates ball release. Gray band represents ±
one standard deviation of the reference normal. This pitcher’s shoulder horizontal abduction far exceeds the
reference normal around footstrike (arrow).
150°
100°
Abd –, Add +
50°
0°
0s 0.2s 0.4s 0.6s 0.8s 1s 1.2s 1.4s

Figure 13.9 Shoulder abduction graph. The first vertical lines indicate footstrike. The second vertical line
indicates ball release. Gray band represents ± one standard deviation of the reference normal. This pitcher is
around 90 degrees of shoulder abduction at footstrike, but then drops his arm as he beings to externally rotate
(arrow). His arm is still low at ball release.
The ideal arm position at footstrike is 90 degrees of abduction, 90 degrees of elbow flexion,
and 60 degrees of shoulder external rotation. The ideal arm position at ball release is 90 degrees of
abduction. The arm should follow a relatively straight path from footstrike to ball release. This
pitcher has 90 degrees of abduction at footstrike (Figure 13.9), but there is a sharp decrease to
40 degrees of shoulder abduction. This is a dramatic deviation from normal and expected values.
At ball release, his shoulder abduction is still lower than anticipated values. Leading with the
elbow and having the elbow lower than the shoulders at ball release will increase the elbow val-
gus torque and the stress on the UCL.
Greater shoulder external rotation has been shown to increase elbow valgus torque (Nicholson
et al., 2021). This pitcher has a maximum shoulder external rotation of 195 degrees when pitch-
ing, while the normal range for collegiate pitchers is 157–180 degrees (Table 13.1). Most pitchers
release the baseball around 100 degrees of shoulder external rotation, but this pitcher released the
ball between 135 and 165 degrees of shoulder external rotation. This indicates he releases the ball
early in the pitch cycle. This reduces internal rotation velocity and can limit pitch velocity. As
with other deficits, the pitcher may compensate with other abnormal mechanics to maintain a
normal and effective velocity during competition.
In summary, this pitcher had abnormalities in sagittal, frontal, and transverse planes in the legs
and trunk. Lower extremity deficits were likely the
ISTUDY
upper extremity. Specifically, he substituted excessive shoulder horizontal adduction and exces-
sive shoulder external rotation to overcome low hip/shoulder separation angles. This led to exces-
sive horizontal abduction motion and increased elbow valgus torque. Inefficient power transfer
and early ball release further contributed to increased elbow stress and inefficient pitching
mechanics. Identifying these flaws and implementing corrections may have been essential in
reducing future injury. Unfortunately, the biomechanical data, although collected, was not eval-
uated and integrated into a training plan and this pitcher sustained a career-ending injury.
13.2.2 References
Agresta, C. E., K. Krieg, and M. T. Freehill. “Risk Factors for Baseball-Related Arm Injuries: A Systematic
Review,” Orthop. J. Sport. Med. 7: 1–13, 2019.
Aguinaldo, A. and R. Escamilla. “Segmental Power Analysis of Sequential Body Motion and Elbow Valgus
Loading During Baseball Pitching: Comparison Between Professional and High School Baseball Players,”
Orthop. J. Sports Med. 7: 1–9, 2019.
Aguinaldo, A. L., J. Buttermore, and H. Chambers. “Effects of Upper Trunk Rotation on Shoulder Joint Tor-
que Among Baseball Pitchers of Various Levels,” J. Appl. Biomech. 23: 42–51, 2007.
Anz, A. W., B. D. Bushnell, L. P. Griffin, T. J. Noonan, M. R. Torry, and R. J. Hawkins. “Correlation of
Torque and Elbow Injury in Professional Baseball Pitchers,” Am. J. Sports Med. 38: 1368–1374, 2010.
Bullock, G. S., J. Strahm, T. C. Hulburt, E. C. Beck, B. R. Waterman, and K. F. Nicholson. “The Relationship
of Range of Motion, Hip Shoulder Separation, and Pitching Kinematics,” Int. J. Sports Phys. Ther. 15:
1119–1128, 2020.
Bullock, G. S., G. Menon, K. Nicholson, R. J. Butler, N. K. Arden, and S. R. Filbay. “Baseball Pitching Bio-
mechanics in Relation to Pain, Injury, and Surgery: A Systematic Review,” J. Sci. Med. Sport 24(1):
13–20, 2020. https://doi.org/10.1016/j.jsams.2020.06.015.
Conte, S. A., C. L. Camp, and J. S. Dines. “Injury Trends in Major League Baseball Over 18 Seasons: 1998-
2015,” Am. J. Orthop. 45: 116–123, 2016.
Dillman, C. I., G. S. Fleisig, and R. Andrews. “Biomechanics of Pitching with Emphasis upon Shoulder
Kinematics,” J. Orthop. Sports Phys. Ther. 18(2): 402–408, 1993.
Feltner, M. and J. Dapena. “Dynamics of the Shoulder and Elbow Joints of the Throwing Arm During a Base-
ball Pitch,” Int. J. Sport Biomech. 2: 235–259, 2016.
Hirashima, M., H. Kadota, S. Sakurai, K. Kudo, and T. Ohtsuki. “Sequential Muscle Activity and Its
Functional Role in the Upper Extremity and Trunk During Overarm Throwing,” J. Sports Sci. 20:
301–310, 2002.
Hootman, J. M., R. Dick, and J. Agel. “Epidemiology of Collegiate Injuries for 15 Sports: Summary and
Recommendations for Injury Prevention Initiatives,” J. Athl. Train. 42: 311–319, 2007.
Nicholson, K. F., T. C. Hulburt, E. C. Beck, B. R. Waterman, and G. S. Bullock. “The Relationship between
Pitch Velocity and Shoulder Distraction Force and Elbow Valgus Torque in Collegiate and High School
Pitchers,” J. Shoulder Elbow Surg. 29(12): 2661–2667, 2020. https://doi.org/10.1016/j.jse.2020.04.046.
Nicholson, K. F., G. S. Collins, B. R. Waterman, and G. S. Bullock. “Machine Learning and Statistical Pre-
diction of Pitching Arm Kinetics,” Am. J. Sports Med. 50(1): 1–10, 2021. https://doi.org/10.1177/
03635465211054506.
Posner, M., K. L. Cameron, J. M. Wolf, P. J. Belmont, and B. D. Owens. “Epidemiology of Major League
Baseball Injuries,” Am. J. Sports Med. 39: 1675–1691, 2011.
Pytiak, A. V., M. J. Kraeutler, D. W. Currie, E. C. McCarty, and R. D. Comstock. “An Epidemiological Com-
parison of Elbow Injuries Among United States High School Baseball and Softball Players, 2005-2006
Through 2014-2015,” Sports Health 10(2): 119–124, 2018. https://doi.org/10.1177/1941738117736493.
Sabick, M. B., M. R. Torry, R. L. Lawton, R. J. Hawkins, and C. Frcs. “Valgus Torque in Youth Baseball
Pitchers: A Biomechanical Study,” J. Shoulder Elbow Surg. 13(3): 349–355, 2004. https://doi.org/
10.1016/j.jse.2004.01.013.
ISTUDY
Sgroi, T., P. N. Chalmers, A. J. Riff, M. Lesniak, E. T. Sayegh, M. A. Wimmer, N. N. Verma, B. J. Cole, and
A. A. Romeo. “Predictors of Throwing Velocity in Youth and Adolescent Pitchers,” J. Shoulder Elbow
Surg. 24: 1339–1345, 2015.
Shanley, E., M. J. Rauh, L. A. Michener, T. S. Ellenbecker, J. C. Garrison, and C. A. Thigpen. “Shoulder
Range of Motion Measures as Risk Factors for Shoulder and Elbow Injuries in High School Softball
and Baseball Players,” Am. J. Sports Med. 39(9): 1997–2006, 2011. https://doi.org/10.1177/
0363546511408876.
Werner, S. L., G. S. Fleisig, C. J. Dillman, and J. R. Andrews. “Biomechanics of the Elbow During Baseball
Pitching,” J. Orthop. Sport. Phys. Ther. 17: 274–278, 1993.
Werner, S. L., T. A. Murray, R. J. Hawkins, and T. J. Gill. “Relationship Between Throwing Mechanics and
Elbow Valgus in Professional Baseball Pitchers,” J. Shoulder Elb. Surg. 11: 151–155, 2002.
ISTUDY
APPENDIX A
KINEMATIC, KINETIC, AND ENERGY DATA
Body mass = 56.7 kg
Frame rate = 69.9 frames/s
Base rib cage
25 cm
Greater trochanter (hip)

cm
.4
31
Lateral epicondyle of thigh (knee)

Head of fibula
42.5 cm
33.6 cm
l
tarsa
eta
cm
Lateral malleolus (ankle) m

th
Fif
.2
12
Heel Toe
TOR, toe off right foot HCR, heel contact right foot
Figure A.1 Walking trial – marker locations and mass and frame rate information.

Companion website:
ISTUDY
TABLE A.1 Raw Coordinate Data (cm)
Base
Time Rib Cage Right Hip Right Knee Right Fibula Right Ankle Right Heel Right Metat. Right Toe
Frame s X Y X Y X Y X Y X Y X Y X Y X Y
1 0.000 46.98 104.41 44.94 78.58 41.00 47.40 35.91 40.53 9.31 21.44 2.95 24.24 7.53 9.35 11.73 3.63
2 0.014 49.22 104.79 47.31 78.58 45.02 46.89 40.06 40.02 12.70 22.46 7.23 26.02 10.54 10.63 13.72 4.77
3 0.029 51.10 105.17 49.57 78.71 48.68 47.27 44.23 40.15 16.49 23.73 10.64 27.30 14.20 12.03 17.63 6.43
4 0.043 53.13 105.30 51.74 79.21 52.50 47.53 48.43 40.15 20.81 24.37 14.71 27.55 18.78 12.53 22.47 6.94
5 0.057 54.86 105.43 53.33 79.09 56.13 47.91 52.70 40.78 24.96 24.24 18.72 27.42 23.17 12.66 27.12 7.32
6 0.072 56.81 106.06 55.41 79.98 59.87 48.67 56.56 41.29 29.33 24.62 23.09 27.17 28.31 12.41 32.51 6.81
7 0.086 58.25 106.32 56.73 80.49 63.34 49.44 60.29 41.80 33.57 23.73 26.95 26.02 33.44 12.03 38.02 6.81
8 0.100 60.03 106.95 58.89 81.00 66.90 50.84 64.36 42.69 38.78 23.22 31.27 25.01 39.55 11.26 44.38 6.30
9 0.114 61.56 107.08 60.79 81.12 69.96 51.09 67.79 43.20 43.11 22.21 35.73 22.84 45.01 10.24 50.36 6.04
10 0.129 63.54 107.46 62.78 82.01 73.22 51.73 71.31 43.84 48.15 21.06 40.64 20.93 51.46 8.97 57.05 5.28
11 0.143 65.20 107.85 64.69 82.40 76.27 53.00 74.74 44.86 53.23 20.04 45.73 19.02 57.56 8.21 63.54 5.15
12 0.157 66.92 107.85 66.67 82.78 79.01 53.51 77.86 45.11 58.27 18.52 51.27 16.86 63.99 7.44 70.23 5.03
13 0.172 68.77 107.59 68.65 82.65 81.62 54.15 80.73 45.49 63.68 16.86 56.56 14.44 70.30 6.30 76.79 4.52
14 0.186 70.37 107.59 70.88 83.16 83.99 54.53 83.48 45.87 69.22 15.59 62.23 12.66 76.73 5.79 83.48 4.77
15 0.200 72.43 107.59 73.19 82.90 86.56 54.78 86.30 46.00 74.72 14.44 68.23 11.01 83.63 5.66 90.12 5.54
16 0.215 74.00 107.59 74.89 83.03 88.51 54.91 88.89 46.13 80.36 13.43 74.13 9.10 89.53 5.28 96.14 5.54
17 0.229 75.68 107.59 76.95 82.90 90.57 54.91 90.95 45.87 85.86 12.66 80.01 7.57 95.53 5.79 102.40 6.55
18 0.243 77.67 107.08 79.19 82.65 93.06 54.65 93.32 46.00 91.54 12.03 86.32 6.55 101.72 5.79 108.84 7.70
19 0.257 79.55 106.95 81.20 82.40 94.56 54.02 95.45 45.62 96.85 11.64 92.02 5.41 107.79 6.17 114.16 8.97
20 0.272 81.47 107.21 83.12 82.14 96.10 54.02 96.99 45.24 101.45 11.77 97.50 5.15 113.28 7.70 119.39 10.88
21 0.286 83.53 106.45 85.69 81.50 98.16 53.25 99.18 44.73 106.56 11.77 103.13 4.77 118.78 8.21 124.38 12.15
22 0.300 85.86 105.68 87.77 80.49 99.73 51.86 101.13 43.84 110.68 11.52 108.39 4.14 123.15 9.35 128.24 13.30
23 0.315 87.74 105.81 89.91 80.61 101.36 51.73 103.14 43.71 114.85 12.15 112.43 4.90 126.94 10.88 131.77 15.33
24 0.329 90.34 105.17 92.25 80.49 103.57 51.09 105.36 43.58 118.33 12.15 116.30 5.28 130.55 11.39 135.39 16.23
25 0.343 92.25 104.79 94.41 80.10 105.23 50.96 107.26 43.46 120.75 12.28 118.84 4.65 133.22 11.64 137.80 16.73
26 0.357 94.36 104.16 97.03 79.72 107.59 50.84 109.50 42.69 122.99 12.03 121.34 4.14 135.21 11.39 139.79 16.23
27 0.372 96.57 103.77 99.37 79.98 109.93 50.58 111.58 42.95 124.31 11.90 122.14 3.75 136.27 10.37 140.85 15.21
28 0.386 98.73 103.52 101.53 79.34 112.34 50.46 113.87 41.80 124.94 10.63 122.65 3.25 137.16 8.97 141.99 13.68
29 0.400 101.40 103.26 104.45 79.60 114.89 50.33 116.28 41.80 125.83 9.61 123.03 3.50 137.66 7.95 142.88 11.90
30 0.415 103.60 103.65 106.78 79.60 116.96 50.46 118.11 41.80 127.02 9.74 123.20 4.14 138.34 6.30 144.45 10.24
ISTUDY
31 0.429 105.94 103.65 108.86 79.60 118.53 50.33 119.81 41.55 127.82 9.48 123.12 4.01 138.77 5.41 145.13 8.59
32 0.443 108.22 103.52 111.28 80.10 120.44 50.58 121.71 42.18 128.33 9.61 123.24 4.14 139.27 4.65 145.38 6.81
33 0.458 110.75 104.03 113.80 80.49 122.84 50.96 124.24 42.44 129.07 9.35 123.60 4.14 139.51 4.14 146.38 6.30
34 0.472 112.91 104.03 115.58 81.38 125.25 51.86 126.14 42.57 129.45 9.61 123.98 4.39 139.63 3.37 146.25 5.41
35 0.486 115.20 104.28 117.36 81.89 127.54 51.86 128.05 42.44 129.83 9.74 123.98 4.26 140.01 3.63 147.14 5.41
36 0.500 116.74 104.54 118.78 81.63 128.32 51.86 128.83 42.69 129.46 9.10 123.48 4.14 139.39 3.75 146.52 4.77
37 0.515 119.05 105.43 120.96 81.63 129.87 51.86 130.13 42.95 129.74 9.35 123.89 4.65 139.80 3.63 146.54 4.90
38 0.529 121.31 105.30 123.10 82.27 130.86 51.60 131.49 42.18 130.10 9.23 124.24 4.52 140.02 3.75 147.27 4.52
39 0.543 123.28 106.19 124.93 83.03 132.18 52.36 132.31 42.82 130.53 9.61 124.55 4.90 140.45 3.75 147.33 5.28
40 0.558 125.03 106.19 126.94 82.78 133.04 51.35 132.79 42.44 130.37 9.35 124.64 4.26 140.55 3.50 147.04 4.77
41 0.572 127.21 106.83 128.61 83.03 133.58 51.73 132.81 42.44 130.27 9.61 124.67 4.77 139.94 3.63 147.19 4.52
42 0.586 128.71 106.95 130.24 83.03 134.31 51.35 133.17 43.20 129.98 8.97 123.75 4.77 139.91 3.37 146.91 4.77
43 0.601 130.90 107.34 132.04 82.90 134.72 50.96 133.70 42.57 130.39 9.61 124.54 4.14 140.19 3.50 147.31 4.52
44 0.615 132.52 107.97 133.03 83.41 135.70 52.62 134.56 43.20 130.23 9.61 124.12 4.52 139.90 4.01 146.78 4.26
45 0.629 134.48 107.85 134.86 83.54 136.26 51.86 134.73 42.95 129.89 9.48 123.91 4.77 139.82 3.63 146.95 4.26
46 0.643 136.33 108.10 136.20 83.80 137.09 51.86 135.18 43.07 130.22 10.24 124.37 4.90 140.02 3.37 147.02 4.77
47 0.658 138.33 108.23 138.08 82.90 137.95 51.47 135.66 42.69 130.06 9.99 124.08 4.77 140.11 3.75 146.99 4.26
48 0.672 140.14 108.48 139.38 82.90 138.74 51.60 136.19 42.69 130.47 10.12 124.36 5.03 140.01 3.25 147.01 4.01
49 0.686 142.20 107.97 140.67 82.78 139.15 51.09 136.73 42.57 130.49 9.99 124.64 5.28 140.16 3.12 147.04 4.52
50 0.701 144.05 107.46 142.02 82.65 140.11 51.09 137.05 42.57 130.56 9.99 124.20 4.90 139.60 3.63 146.85 4.01
51 0.715 146.01 107.08 143.85 82.01 140.92 50.58 137.87 42.06 130.49 10.24 124.76 4.77 140.03 3.50 147.28 3.75
52 0.729 148.58 107.46 146.04 82.40 142.22 50.84 138.66 42.44 131.27 10.24 125.17 5.28 140.18 3.63 147.18 4.14
53 0.744 149.92 106.57 147.51 81.63 143.05 50.58 139.23 41.93 130.96 10.12 125.24 5.03 140.25 3.37 147.25 4.01
54 0.758 152.34 106.32 148.90 81.50 143.69 50.71 140.12 42.57 131.47 10.63 125.49 5.66 140.50 3.37 147.50 3.88
55 0.772 153.94 105.81 150.50 81.12 144.27 50.46 140.45 42.95 131.03 10.63 124.80 6.17 140.07 2.99 146.94 3.75
56 0.786 156.15 105.30 152.46 81.00 145.59 50.71 141.52 42.31 131.34 10.63 125.61 5.92 140.50 3.37 147.24 3.88
57 0.801 158.65 104.92 154.57 81.00 147.07 50.58 142.61 42.18 131.80 10.75 125.94 7.06 140.70 3.50 147.45 3.75
58 0.815 160.50 104.66 156.30 80.61 148.41 50.71 143.83 43.20 132.12 12.15 125.89 7.70 140.65 4.01 147.39 4.14
59 0.829 162.61 104.41 157.90 80.10 149.75 50.33 144.66 42.31 131.94 12.03 125.96 7.83 140.34 3.75 146.95 3.75
60 0.844 164.58 104.03 159.88 80.23 151.22 50.33 146.26 42.82 132.26 12.66 126.03 8.46 140.41 3.75 147.02 3.88
61 0.858 166.66 103.77 162.08 80.23 153.30 50.33 147.96 42.69 132.69 12.92 126.58 9.61 140.20 4.26 146.69 4.26
62 0.872 168.90 104.03 164.19 79.60 155.41 50.07 150.06 42.57 133.39 13.93 126.90 10.50 140.01 3.88 146.76 4.01
63 0.887 171.00 103.90 166.42 79.85 157.01 50.07 152.30 42.69 133.97 14.83 127.48 12.28 140.21 3.88 146.83 3.75
64 0.901 173.53 103.14 169.07 78.96 160.29 49.06 154.95 42.06 135.48 14.95 128.73 13.04 140.95 4.14 147.06 3.50
(continued )
ISTUDY
TABLE A.1 (Continued)
Base
Time Rib Cage Right Hip Right Knee Right Fibula Right Ankle Right Heel Right Metat. Right Toe
Frame s X Y X Y X Y X Y X Y X Y X Y X Y
65 0.915 175.81 103.52 171.74 79.47 163.22 49.06 157.62 42.31 136.75 16.48 129.62 14.83 141.08 4.65 147.18 3.75
66 0.929 179.23 103.14 175.16 78.83 166.89 48.55 161.03 41.04 139.14 17.37 132.15 16.10 141.94 5.28 148.31 2.99
67 0.944 181.49 103.14 177.80 78.58 170.04 48.16 164.69 41.42 140.90 18.64 134.02 18.52 142.42 5.79 148.40 2.86
68 0.958 183.65 103.77 180.60 79.09 174.11 48.04 168.25 41.55 143.69 19.79 136.95 21.57 143.69 7.57 149.04 3.63
69 0.972 184.86 103.39 182.44 77.94 176.97 47.02 171.50 40.15 145.54 20.42 139.43 23.35 144.14 7.95 148.59 3.50
70 0.987 187.71 104.28 185.67 78.83 181.73 47.53 176.00 40.15 149.53 21.82 143.81 25.13 146.86 9.74 150.17 4.26
71 1.001 188.82 104.16 187.29 78.20 184.36 46.89 179.40 39.51 152.68 22.21 146.95 25.90 149.11 10.75 151.91 4.14
72 1.015 191.13 104.92 189.48 78.83 188.33 47.27 184.00 39.89 156.39 23.73 150.54 27.17 153.34 12.03 156.52 5.66
73 1.030 192.42 105.55 190.89 79.09 191.78 47.66 187.71 40.53 159.84 24.24 153.73 27.68 157.17 12.28 160.73 5.92
74 1.044 194.40 106.19 193.00 79.60 195.29 48.16 191.47 41.04 163.99 24.37 157.50 27.93 161.82 12.66 165.38 6.43
75 1.058 196.18 105.94 195.16 79.72 199.10 48.93 195.79 41.42 168.44 24.50 161.95 27.04 167.29 12.15 170.98 6.68
76 1.072 197.69 106.06 196.42 80.10 202.66 49.31 199.47 41.55 172.50 23.61 165.88 25.77 171.99 10.88 176.57 6.17
77 1.087 199.57 106.70 198.68 81.00 206.19 50.33 203.39 42.44 177.30 22.84 170.56 25.01 177.81 10.12 182.65 6.17
78 1.101 201.17 107.21 200.53 81.63 209.57 51.22 206.90 43.20 182.08 21.95 175.08 23.10 183.74 9.61 189.21 4.90
79 1.115 202.99 107.59 202.48 82.14 212.66 52.11 210.63 43.84 187.09 21.06 179.83 21.57 189.76 8.97 195.61 4.77
80 1.130 204.67 107.59 204.80 82.65 215.74 52.62 213.83 44.47 192.20 19.66 184.95 19.28 196.02 7.19 201.75 4.39
81 1.144 206.27 107.46 206.78 82.52 218.49 53.13 216.96 44.47 197.36 17.88 190.11 16.73 202.45 6.81 208.94 3.75
82 1.158 208.30 107.72 209.07 83.03 221.28 53.89 219.88 45.24 202.96 16.86 195.58 14.57 209.19 5.41 215.94 4.26
83 1.173 210.18 107.72 211.07 83.03 223.80 54.02 222.91 45.24 208.15 15.59 201.27 12.53 215.14 4.90 222.65 3.50
84 1.187 211.95 107.46 212.97 82.78 226.20 54.40 225.70 45.49 213.86 13.68 207.37 10.12 221.88 4.39 229.13 3.88
85 1.201 213.83 107.21 215.10 82.90 228.59 54.65 228.21 45.49 219.43 12.79 213.45 8.59 228.72 3.75 235.59 4.90
86 1.215 215.63 107.08 217.03 82.65 230.52 54.65 230.52 45.11 224.92 11.39 219.32 6.68 234.84 4.26 241.72 5.28
87 1.230 217.58 106.70 219.11 82.65 232.85 54.65 232.60 45.24 230.94 11.39 225.85 5.66 241.38 4.52 247.74 6.30
88 1.244 219.36 106.45 220.89 82.27 234.63 53.64 234.76 44.47 236.28 11.01 232.09 4.52 247.86 4.52 254.10 7.32
89 1.258 221.06 105.94 222.71 81.63 236.07 53.13 236.84 44.35 241.80 10.37 237.35 4.01 253.00 6.05 259.36 9.48
90 1.273 223.53 105.55 225.18 81.12 238.29 52.75 239.18 43.46 247.07 10.63 243.89 3.63 259.41 7.06 265.01 11.26
91 1.287 225.31 104.92 227.60 81.00 239.81 51.73 241.47 43.33 252.03 11.01 249.23 3.63 264.24 8.84 269.33 12.92
92 1.301 227.33 104.66 229.75 80.10 241.33 50.84 243.36 42.82 255.83 11.01 253.54 3.63 268.31 9.86 273.40 14.32
93 1.316 229.28 104.41 231.57 79.98 243.03 50.46 245.32 42.82 259.31 11.26 257.53 4.14 271.78 10.75 276.49 15.84
94 1.330 231.49 104.03 234.16 79.60 245.23 50.33 247.40 42.82 262.03 12.15 260.50 4.01 274.75 11.26 278.95 16.48
ISTUDY
95 1.344 233.80 103.52 236.47 79.21 247.67 49.82 249.57 42.31 263.95 10.88 262.55 3.63 276.42 10.88 281.39 15.72
96 1.358 235.62 103.01 238.80 78.83 250.00 49.82 251.91 41.68 265.27 10.24 263.49 3.12 277.74 9.99 282.70 14.70
97 1.373 238.04 103.01 241.60 79.21 252.80 50.33 254.32 42.06 266.41 9.86 264.12 2.99 278.63 8.59 283.72 13.43
98 1.387 240.34 103.14 244.29 79.21 255.23 50.20 256.50 41.55 267.06 9.23 264.14 3.37 279.15 7.44 284.50 11.39
99 1.401 242.88 103.01 246.83 79.09 257.39 50.07 258.92 41.42 268.72 9.48 264.52 3.63 280.42 6.05 286.15 9.10
100 1.416 245.09 102.63 249.16 79.34 259.60 50.07 260.74 41.68 269.78 9.23 264.69 3.63 281.10 4.90 287.08 7.95
101 1.430 247.53 103.26 251.60 79.72 261.90 50.96 263.30 42.57 270.18 9.48 265.21 4.26 281.25 4.39 287.86 6.68
102 1.444 249.91 102.88 253.73 80.36 264.42 50.96 265.95 42.44 270.78 9.10 265.18 3.63 281.47 3.50 288.09 5.54
103 1.459 251.81 103.39 255.63 81.00 266.70 51.60 267.59 42.82 271.15 9.23 265.30 4.39 281.84 3.63 288.20 5.15
104 1.473 253.96 104.16 257.65 81.38 268.47 51.47 269.23 42.69 271.40 9.23 265.16 3.88 281.96 3.25 288.45 5.03
105 1.487 256.07 104.28 259.51 81.25 269.56 51.35 270.19 42.31 271.34 8.59 265.36 3.50 281.52 2.74 288.14 4.39
106 1.501 257.61 105.05 260.92 81.63 270.34 51.60 270.85 42.44 270.85 8.84 264.99 3.88 281.15 2.35 287.77 4.52
ISTUDY
TABLE A.2a Filtered Marker Kinematics – Rib Cage and Greater Trochanter (Hip)
Base Rib Cage Right Hip
Frame Time X VX AX Y VY AY X VX AX Y VY AY
s m m/s m/s/s m m/s m/s/s m m/s m/s/s m m/s m/s/s
TOR 1 0.000 0.4695 1.43 0.3 1.0435 0.22 1.0 0.4474 1.64 −2.2 0.7870 0.01 3.3
2 0.014 0.4900 1.41 −1.6 1.0467 0.23 0.5 0.4706 1.59 −4.6 0.7875 0.07 4.3
3 0.029 0.5100 1.38 −2.9 1.0501 0.24 0.4 0.4928 1.51 −5.6 0.7889 0.14 4.8
4 0.043 0.5294 1.33 −3.4 1.0535 0.24 0.6 0.5138 1.43 −5.3 0.7914 0.21 4.7
5 0.057 0.5481 1.28 −3.3 1.0570 0.25 0.6 0.5336 1.36 −3.9 0.7948 0.27 3.9
6 0.072 0.5661 1.24 −2.5 1.0607 0.26 0.2 0.5526 1.31 −2.0 0.7991 0.32 2.6
7 0.086 0.5836 1.21 −1.5 1.0644 0.26 −0.7 0.5712 1.30 −0.2 0.8039 0.34 1.2
8 0.100 0.6007 1.20 −0.5 1.0680 0.24 −1.8 0.5898 1.31 1.0 0.8089 0.35 −0.2
9 0.114 0.6178 1.20 0.2 1.0713 0.21 −2.8 0.6086 1.33 1.7 0.8139 0.34 −1.5
10 0.129 0.6349 1.20 0.5 1.0739 0.16 −3.5 0.6278 1.36 1.9 0.8185 0.31 −2.8
11 0.143 0.6521 1.21 0.5 1.0759 0.11 −3.7 0.6474 1.38 1.9 0.8226 0.26 −3.8
12 0.157 0.6695 1.22 0.5 1.0770 0.05 −3.4 0.6673 1.41 1.6 0.8259 0.20 −4.4
13 0.172 0.6869 1.22 0.5 1.0774 0.01 −2.7 0.6876 1.43 1.1 0.8283 0.13 −4.6
14 0.186 0.7044 1.23 0.6 1.0773 −0.02 −2.1 0.7082 1.44 0.5 0.8298 0.07 −4.7
15 0.200 0.7221 1.24 1.0 1.0767 −0.05 −1.9 0.7288 1.44 0.2 0.8302 0.00 −4.7
16 0.215 0.7400 1.26 1.6 1.0758 −0.08 −2.1 0.7495 1.45 0.3 0.8298 −0.07 −4.6
17 0.229 0.7581 1.29 2.2 1.0745 −0.11 −2.3 0.7702 1.45 0.7 0.8283 −0.13 −4.4
18 0.243 0.7768 1.32 2.8 1.0727 −0.14 −2.5 0.7910 1.47 1.1 0.8260 −0.19 −4.0
19 0.257 0.7960 1.37 3.0 1.0704 −0.18 −2.8 0.8121 1.48 1.4 0.8228 −0.25 −3.1
20 0.272 0.8158 1.41 2.9 1.0675 −0.22 −2.8 0.8335 1.51 1.7 0.8189 −0.28 −1.7
21 0.286 0.8363 1.45 2.5 1.0640 −0.26 −2.4 0.8552 1.53 1.8 0.8147 −0.30 0.1
22 0.300 0.8573 1.48 2.0 1.0600 −0.29 −1.6 0.8773 1.56 1.9 0.8105 −0.28 1.6
23 0.315 0.8787 1.51 1.4 1.0557 −0.31 −0.7 0.8997 1.59 2.1 0.8067 −0.25 2.3
24 0.329 0.9003 1.52 1.1 1.0512 −0.31 0.3 0.9226 1.62 2.2 0.8033 −0.21 2.6
25 0.343 0.9222 1.54 1.3 1.0468 −0.30 1.7 0.9460 1.65 2.1 0.8005 −0.18 2.8
26 0.357 0.9443 1.56 1.6 1.0427 −0.26 3.2 0.9698 1.68 1.6 0.7983 −0.13 3.1
27 0.372 0.9668 1.58 1.8 1.0392 −0.21 4.4 0.9940 1.69 0.8 0.7967 −0.09 3.7
HCR 28 0.386 0.9896 1.61 1.5 1.0367 −0.14 5.0 1.0183 1.70 −0.4 0.7959 −0.03 4.4
29 0.400 1.0128 1.63 0.9 1.0353 −0.06 4.9 1.0425 1.68 −1.7 0.7959 0.04 4.9
ISTUDY
30 0.415 1.0362 1.64 0.0 1.0349 0.00 4.3 1.0664 1.65 −3.0 0.7970 0.11 5.0
31 0.429 1.0596 1.63 −0.9 1.0353 0.06 3.7 1.0897 1.60 −4.0 0.7991 0.18 4.5
32 0.443 1.0828 1.61 −1.8 1.0366 0.11 3.2 1.1121 1.53 −4.7 0.8023 0.24 3.1
33 0.458 1.1056 1.58 −2.5 1.0384 0.15 2.9 1.1336 1.46 −4.7 0.8060 0.27 1.1
34 0.472 1.1279 1.54 −2.8 1.0408 0.19 2.7 1.1540 1.40 −3.8 0.8100 0.27 −0.6
35 0.486 1.1496 1.50 −2.6 1.0438 0.23 2.3 1.1736 1.36 −2.3 0.8139 0.25 −1.5
36 0.500 1.1707 1.46 −2.2 1.0473 0.26 1.8 1.1928 1.34 −1.0 0.8173 0.23 −1.4
37 0.515 1.1915 1.43 −2.0 1.0512 0.28 1.0 1.2118 1.33 −0.8 0.8205 0.21 −1.3
38 0.529 1.2118 1.41 −2.1 1.0552 0.29 0.3 1.2307 1.31 −1.4 0.8234 0.19 −1.6
39 0.543 1.2317 1.38 −2.1 1.0594 0.29 −0.2 1.2494 1.29 −2.4 0.8260 0.17 −2.0
40 0.558 1.2511 1.35 −1.8 1.0634 0.28 −0.7 1.2675 1.25 −3.2 0.8282 0.14 −2.0
41 0.572 1.2702 1.32 −1.3 1.0673 0.27 −1.1 1.2850 1.20 −3.6 0.8300 0.11 −1.7
42 0.586 1.2890 1.31 −0.7 1.0710 0.25 −1.6 1.3018 1.14 −3.3 0.8314 0.09 −1.6
43 0.601 1.3076 1.30 −0.2 1.0744 0.22 −2.4 1.3178 1.10 −2.6 0.8325 0.07 −1.9
44 0.615 1.3263 1.30 0.3 1.0773 0.18 −3.2 1.3332 1.07 −1.7 0.8333 0.03 −2.7
45 0.629 1.3450 1.31 0.7 1.0796 0.13 −4.0 1.3484 1.05 −0.9 0.8335 −0.01 −3.3
46 0.643 1.3638 1.33 1.0 1.0810 0.07 −4.6 1.3633 1.04 −0.3 0.8330 −0.06 −3.3
47 0.658 1.3829 1.34 1.1 1.0815 0.00 −5.0 1.3782 1.04 0.3 0.8318 −0.11 −2.6
48 0.672 1.4021 1.36 1.1 1.0809 −0.08 −4.8 1.3931 1.05 1.1 0.8300 −0.14 −1.8
49 0.686 1.4217 1.37 1.1 1.0793 −0.14 −4.0 1.4083 1.07 1.9 0.8279 −0.16 −1.2
50 0.701 1.4414 1.39 0.9 1.0769 −0.19 −3.0 1.4239 1.11 2.2 0.8255 −0.17 −0.8
51 0.715 1.4614 1.40 0.6 1.0739 −0.23 −2.3 1.4400 1.14 1.8 0.8230 −0.18 −0.5
52 0.729 1.4814 1.40 0.4 1.0704 −0.26 −1.9 1.4564 1.16 1.2 0.8204 −0.18 −0.2
53 0.744 1.5015 1.41 0.5 1.0666 −0.28 −1.3 1.4731 1.17 1.1 0.8178 −0.18 0.2
54 0.758 1.5217 1.42 0.8 1.0624 −0.29 −0.5 1.4900 1.19 1.4 0.8151 −0.18 0.5
55 0.772 1.5421 1.43 0.9 1.0582 −0.29 0.5 1.5071 1.21 1.8 0.8127 −0.17 0.6
56 0.786 1.5626 1.44 0.7 1.0541 −0.28 1.3 1.5247 1.24 1.9 0.8103 −0.16 0.5
57 0.801 1.5833 1.45 0.5 1.0503 −0.25 1.9 1.5426 1.27 2.3 0.8081 −0.15 0.4
58 0.815 1.6041 1.46 0.7 1.0468 −0.22 2.1 1.5610 1.31 3.1 0.8059 −0.15 0.4
59 0.829 1.6250 1.47 1.5 1.0439 −0.19 2.1 1.5800 1.36 4.5 0.8038 −0.14 0.3
60 0.844 1.6461 1.50 2.6 1.0413 −0.16 1.9 1.5999 1.44 6.0 0.8018 −0.14 0.0
61 0.858 1.6678 1.54 3.6 1.0392 −0.14 1.5 1.6210 1.53 7.2 0.7998 −0.14 −0.3
62 0.872 1.6903 1.60 4.0 1.0373 −0.12 1.2 1.6436 1.64 7.7 0.7977 −0.15 −0.3
63 0.887 1.7136 1.66 3.6 1.0357 −0.10 1.5 1.6679 1.75 7.3 0.7955 −0.15 −0.1
(continued )
ISTUDY
TABLE A.2a (Continued)
Base Rib Cage Right Hip
64 0.901 1.7377 1.70 1.8 1.0343 −0.08 2.3 1.6937 1.85 5.5 0.7933 −0.15 0.2
65 0.915 1.7623 1.71 −1.0 1.0334 −0.04 3.2 1.7208 1.91 2.5 0.7911 −0.15 0.7
66 0.929 1.7866 1.67 −4.0 1.0332 0.01 4.0 1.7483 1.92 −1.1 0.7891 −0.13 1.3
67 0.944 1.8101 1.60 −6.0 1.0338 0.07 4.3 1.7756 1.88 −4.2 0.7873 −0.11 2.0
68 0.958 1.8323 1.50 −6.4 1.0353 0.14 4.2 1.8020 1.80 −6.4 0.7859 −0.08 2.9
69 0.972 1.8531 1.41 −5.8 1.0377 0.19 3.8 1.8270 1.69 −7.5 0.7851 −0.03 3.8
TOR 70 0.987 1.8727 1.34 −4.7 1.0409 0.24 3.0 1.8505 1.58 −7.7 0.7851 0.03 4.5
71 1.001 1.8914 1.28 −3.5 1.0447 0.28 1.8 1.8723 1.48 −6.8 0.7861 0.10 4.8
72 1.015 1.9093 1.24 −2.4 1.0489 0.30 0.4 1.8927 1.39 −5.0 0.7880 0.17 4.6
73 1.030 1.9268 1.21 −1.4 1.0532 0.29 −0.8 1.9120 1.33 −3.0 0.7909 0.23 4.1
74 1.044 1.9440 1.20 −0.7 1.0572 0.27 −1.3 1.9308 1.30 −1.1 0.7947 0.29 3.5
75 1.058 1.9610 1.19 −0.3 1.0610 0.26 −1.1 1.9493 1.30 0.5 0.7991 0.33 2.8
76 1.072 1.9780 1.19 0.1 1.0645 0.24 −1.1 1.9680 1.32 1.7 0.8042 0.37 1.5
77 1.087 1.9951 1.19 0.4 1.0679 0.22 −1.6 1.9870 1.35 2.4 0.8096 0.38 −0.4
78 1.101 2.0122 1.20 0.7 1.0710 0.19 −2.6 2.0066 1.39 2.5 0.8149 0.36 −2.3
79 1.115 2.0294 1.21 1.0 1.0735 0.15 −3.4 2.0267 1.42 2.0 0.8198 0.31 −3.9
80 1.130 2.0469 1.23 1.3 1.0753 0.10 −3.7 2.0473 1.45 1.1 0.8238 0.24 −4.8
81 1.144 2.0647 1.25 1.3 1.0763 0.04 −3.8 2.0681 1.45 0.1 0.8268 0.17 −5.0
82 1.158 2.0827 1.27 1.0 1.0765 −0.01 −3.7 2.0889 1.45 −0.8 0.8287 0.10 −4.9
83 1.173 2.1009 1.28 0.7 1.0760 −0.06 −3.6 2.1095 1.43 −1.1 0.8297 0.03 −4.7
84 1.187 2.1193 1.29 0.5 1.0748 −0.11 −3.4 2.1298 1.42 −1.1 0.8296 −0.03 −4.4
85 1.201 2.1378 1.30 0.6 1.0728 −0.16 −3.0 2.1500 1.40 −0.7 0.8287 −0.09 −4.2
86 1.215 2.1563 1.30 0.8 1.0702 −0.20 −2.7 2.1699 1.40 0.1 0.8270 −0.15 −4.1
87 1.230 2.1751 1.32 1.1 1.0671 −0.24 −2.3 2.1899 1.40 1.0 0.8244 −0.21 −3.8
88 1.244 2.1940 1.34 1.5 1.0635 −0.27 −1.8 2.2100 1.43 2.0 0.8210 −0.26 −3.1
89 1.258 2.2133 1.36 1.6 1.0595 −0.29 −1.1 2.2306 1.46 2.6 0.8169 −0.30 −2.0
90 1.273 2.2329 1.38 1.6 1.0552 −0.30 −0.3 2.2518 1.50 2.6 0.8124 −0.32 −0.7
91 1.287 2.2529 1.41 1.7 1.0510 −0.30 0.4 2.2735 1.54 2.6 0.8078 −0.32 0.6
92 1.301 2.2731 1.43 1.9 1.0468 −0.29 0.8 2.2957 1.57 2.8 0.8033 −0.30 1.8
93 1.316 2.2938 1.46 2.2 1.0428 −0.27 1.3 2.3185 1.62 3.3 0.7992 −0.27 3.0
ISTUDY
94 1.330 2.3149 1.49 2.4 1.0390 −0.25 1.9 2.3419 1.67 3.5 0.7957 −0.22 4.0
95 1.344 2.3365 1.53 2.6 1.0356 −0.22 2.8 2.3661 1.72 3.1 0.7930 −0.15 4.8
96 1.358 2.3587 1.57 2.6 1.0328 −0.17 3.5 2.3910 1.75 2.0 0.7913 −0.08 5.3
HCR 97 1.373 2.3813 1.60 2.2 1.0307 −0.12 3.7 2.4163 1.77 0.2 0.7908 0.00 5.3
98 1.387 2.4045 1.63 1.3 1.0294 −0.06 3.9 2.4417 1.76 −1.7 0.7913 0.07 5.2
99 1.401 2.4279 1.64 0.1 1.0289 −0.01 4.1 2.4667 1.72 −3.5 0.7929 0.15 4.8
100 1.416 2.4514 1.63 −1.1 1.0292 0.05 4.5 2.4909 1.66 −4.7 0.7955 0.21 4.0
101 1.430 2.4746 1.61 −2.3 1.0304 0.12 4.7 2.5142 1.59 −5.2 0.7989 0.26 2.4
102 1.444 2.4974 1.57 −3.2 1.0327 0.19 4.4 2.5364 1.51 −5.3 0.8029 0.28 0.6
103 1.459 2.5194 1.52 −4.0 1.0358 0.25 3.5 2.5575 1.44 −5.2 0.8069 0.28 −1.0
104 1.473 2.5407 1.45 −4.9 1.0398 0.29 2.1 2.5775 1.36 −5.2 0.8108 0.25 −1.8
105 1.487 2.5610 1.37 −6.4 1.0441 0.31 0.4 2.5965 1.29 −6.0 0.8142 0.22 −2.0
106 1.501 2.5800 1.27 −8.6 1.0485 0.30 −1.1 2.6143 1.19 −7.9 0.8171 0.19 −2.1
ISTUDY
TABLE A.2b Filtered Marker Kinematics – Femoral Lateral Epicondyle (Knee) and Head of Fibula
Right Knee Right Fibula
TOR 1 0.000 0.4075 2.66 6.0 0.4754 −0.13 5.9 0.3570 2.84 8.4 0.4052 −0.16 5.1
2 0.014 0.4461 2.71 1.6 0.4741 −0.03 8.3 0.3985 2.93 3.7 0.4036 −0.07 7.1
3 0.029 0.4850 2.71 −1.5 0.4746 0.10 9.5 0.4407 2.95 0.0 0.4034 0.05 8.2
4 0.043 0.5235 2.67 −3.5 0.4771 0.24 9.4 0.4829 2.93 −2.8 0.4049 0.17 8.2
5 0.057 0.5613 2.61 −4.8 0.4815 0.37 8.2 0.5244 2.87 −4.8 0.4082 0.28 7.1
6 0.072 0.5980 2.53 −5.8 0.4877 0.48 6.1 0.5649 2.79 −5.9 0.4130 0.37 5.4
7 0.086 0.6336 2.44 −6.5 0.4952 0.55 3.6 0.6041 2.70 −6.7 0.4189 0.44 3.4
8 0.100 0.6678 2.34 −7.0 0.5034 0.58 1.2 0.6420 2.60 −7.2 0.4255 0.47 1.4
9 0.114 0.7006 2.24 −7.4 0.5118 0.58 −0.7 0.6784 2.49 −7.7 0.4324 0.48 −0.7
10 0.129 0.7319 2.13 −7.8 0.5200 0.56 −2.4 0.7133 2.38 −8.1 0.4391 0.45 −2.5
11 0.143 0.7615 2.02 −8.0 0.5278 0.51 −4.1 0.7464 2.26 −8.3 0.4453 0.40 −4.1
12 0.157 0.7896 1.90 −7.9 0.5347 0.44 −5.7 0.7779 2.14 −8.4 0.4506 0.34 −5.1
13 0.172 0.8159 1.79 −7.6 0.5404 0.35 −7.0 0.8076 2.02 −8.3 0.4549 0.26 −5.7
14 0.186 0.8407 1.68 −7.2 0.5447 0.24 −7.8 0.8357 1.90 −8.1 0.4580 0.17 −6.0
15 0.200 0.8641 1.58 −6.7 0.5474 0.13 −8.3 0.8620 1.79 −7.9 0.4598 0.08 −6.1
16 0.215 0.8860 1.49 −6.2 0.5483 0.01 −8.3 0.8868 1.68 −7.4 0.4604 0.00 −5.9
17 0.229 0.9067 1.41 −5.6 0.5476 −0.11 −8.0 0.9100 1.57 −6.5 0.4597 −0.09 −5.6
18 0.243 0.9263 1.33 −4.7 0.5451 −0.22 −7.1 0.9318 1.49 −5.1 0.4579 −0.16 −5.0
19 0.257 0.9448 1.27 −3.2 0.5412 −0.31 −5.6 0.9526 1.43 −3.3 0.4551 −0.23 −3.9
20 0.272 0.9627 1.24 −1.2 0.5361 −0.38 −3.6 0.9727 1.40 −1.4 0.4514 −0.27 −2.4
21 0.286 0.9803 1.24 0.9 0.5303 −0.42 −0.9 0.9925 1.39 0.2 0.4472 −0.30 −0.6
22 0.300 0.9981 1.27 2.8 0.5242 −0.41 1.8 1.0124 1.40 1.2 0.4430 −0.29 0.6
23 0.315 1.0165 1.32 4.2 0.5186 −0.36 3.8 1.0326 1.42 1.7 0.4389 −0.28 1.1
24 0.329 1.0358 1.39 5.0 0.5138 −0.30 4.7 1.0531 1.45 1.8 0.4350 −0.26 1.0
25 0.343 1.0561 1.46 4.8 0.5100 −0.23 4.7 1.0741 1.48 1.6 0.4314 −0.25 1.0
26 0.357 1.0776 1.52 3.7 0.5072 −0.17 4.4 1.0953 1.50 1.0 0.4279 −0.23 1.5
27 0.372 1.0997 1.57 1.8 0.5053 −0.10 4.2 1.1169 1.51 0.2 0.4247 −0.21 2.6
HCR 28 0.386 1.1223 1.58 −0.2 0.5042 −0.05 4.1 1.1384 1.50 −0.8 0.4220 −0.16 4.0
29 0.400 1.1448 1.56 −1.6 0.5039 0.01 4.2 1.1598 1.48 −1.5 0.4202 −0.09 4.9
30 0.415 1.1669 1.53 −2.1 0.5046 0.07 4.1 1.1808 1.46 −1.9 0.4194 −0.02 4.8
ISTUDY
31 0.429 1.1885 1.50 −2.3 0.5061 0.13 3.5 1.2015 1.43 −2.6 0.4196 0.05 3.7
32 0.443 1.2097 1.46 −3.2 0.5083 0.17 2.0 1.2216 1.38 −4.0 0.4207 0.09 1.9
33 0.458 1.2304 1.41 −5.3 0.5110 0.19 −0.1 1.2410 1.31 −6.2 0.4221 0.10 0.1
34 0.472 1.2500 1.31 −7.7 0.5138 0.17 −2.1 1.2592 1.21 −8.5 0.4235 0.09 −1.1
35 0.486 1.2679 1.19 −9.5 0.5160 0.13 −3.3 1.2755 1.07 −9.9 0.4247 0.07 −1.7
36 0.500 1.2839 1.04 −9.8 0.5175 0.08 −3.5 1.2898 0.92 −10.4 0.4255 0.04 −1.6
37 0.515 1.2977 0.90 −9.3 0.5182 0.03 −2.9 1.3019 0.77 −10.4 0.4260 0.02 −1.0
38 0.529 1.3097 0.78 −8.2 0.5183 −0.01 −2.0 1.3119 0.63 −9.8 0.4262 0.02 −0.1
39 0.543 1.3200 0.67 −7.0 0.5180 −0.03 −0.9 1.3198 0.49 −8.3 0.4265 0.02 0.6
40 0.558 1.3288 0.58 −5.4 0.5175 −0.03 0.3 1.3260 0.39 −6.0 0.4269 0.03 0.8
41 0.572 1.3365 0.51 −3.4 0.5171 −0.02 1.3 1.3309 0.32 −3.3 0.4274 0.04 0.4
42 0.586 1.3435 0.48 −1.5 0.5170 0.00 1.3 1.3352 0.29 −1.2 0.4281 0.05 −0.3
43 0.601 1.3502 0.47 −0.1 0.5172 0.02 0.2 1.3393 0.29 0.0 0.4287 0.04 −1.1
44 0.615 1.3570 0.48 0.7 0.5175 0.01 −1.4 1.3435 0.29 0.5 0.4291 0.01 −1.8
45 0.629 1.3639 0.49 1.0 0.5175 −0.02 −2.5 1.3477 0.30 0.8 0.4291 −0.02 −2.0
46 0.643 1.3710 0.51 1.0 0.5169 −0.06 −2.6 1.3521 0.32 1.1 0.4287 −0.04 −1.8
47 0.658 1.3784 0.52 1.0 0.5157 −0.10 −2.0 1.3567 0.33 1.4 0.4278 −0.07 −1.1
48 0.672 1.3859 0.54 1.2 0.5141 −0.12 −1.0 1.3617 0.35 1.6 0.4268 −0.08 −0.3
49 0.686 1.3937 0.55 1.4 0.5123 −0.13 0.1 1.3669 0.38 1.8 0.4256 −0.08 0.5
50 0.701 1.4018 0.58 1.4 0.5105 −0.12 1.1 1.3725 0.41 1.9 0.4246 −0.06 1.3
51 0.715 1.4102 0.59 1.1 0.5089 −0.10 1.7 1.3785 0.43 1.8 0.4238 −0.04 1.9
52 0.729 1.4188 0.61 1.1 0.5077 −0.07 1.8 1.3849 0.46 1.8 0.4235 −0.01 2.1
53 0.744 1.4275 0.62 1.8 0.5069 −0.04 1.6 1.3917 0.49 2.2 0.4236 0.02 1.6
54 0.758 1.4366 0.66 3.2 0.5065 −0.02 1.1 1.3988 0.52 3.0 0.4241 0.04 0.8
55 0.772 1.4463 0.72 4.7 0.5062 −0.01 0.5 1.4066 0.57 4.1 0.4247 0.04 0.2
56 0.786 1.4571 0.79 5.7 0.5061 −0.01 −0.3 1.4151 0.64 5.4 0.4254 0.04 −0.1
57 0.801 1.4690 0.88 6.5 0.5059 −0.02 −1.0 1.4248 0.73 6.8 0.4260 0.04 −0.4
58 0.815 1.4822 0.98 7.5 0.5055 −0.04 −1.5 1.4359 0.83 8.5 0.4265 0.03 −0.9
59 0.829 1.4969 1.09 8.9 0.5048 −0.06 −2.0 1.4486 0.97 10.4 0.4269 0.02 −1.5
60 0.844 1.5135 1.23 10.6 0.5037 −0.10 −2.5 1.4635 1.13 12.1 0.4270 −0.01 −2.1
61 0.858 1.5322 1.40 12.3 0.5021 −0.14 −3.1 1.4809 1.31 13.4 0.4267 −0.04 −2.7
62 0.872 1.5534 1.58 13.8 0.4999 −0.18 −3.2 1.5011 1.51 14.2 0.4257 −0.09 −3.2
63 0.887 1.5775 1.79 14.5 0.4969 −0.23 −2.8 1.5242 1.72 14.4 0.4242 −0.14 −3.3
64 0.901 1.6046 2.00 13.8 0.4933 −0.26 −1.9 1.5503 1.93 14.1 0.4219 −0.18 −2.9
(continued )
ISTUDY
TABLE A.2b (Continued)
Right Knee Right Fibula
65 0.915 1.6346 2.19 11.7 0.4894 −0.28 −0.8 1.5793 2.12 13.0 0.4189 −0.22 −2.2
66 0.929 1.6671 2.33 9.0 0.4853 −0.28 0.5 1.6110 2.30 11.4 0.4156 −0.24 −1.2
67 0.944 1.7014 2.44 6.1 0.4812 −0.27 2.0 1.6450 2.45 9.4 0.4119 −0.26 −0.1
68 0.958 1.7369 2.51 3.5 0.4776 −0.23 3.8 1.6810 2.57 7.6 0.4083 −0.25 1.8
69 0.972 1.7732 2.54 1.6 0.4747 −0.16 5.6 1.7185 2.66 5.9 0.4049 −0.20 4.3
TOR 70 0.987 1.8097 2.55 0.3 0.4730 −0.07 7.1 1.7572 2.74 4.1 0.4024 −0.12 6.8
71 1.001 1.8462 2.55 −0.3 0.4728 0.04 8.1 1.7967 2.78 2.3 0.4014 −0.01 8.2
72 1.015 1.8826 2.54 −0.5 0.4743 0.16 8.2 1.8368 2.80 0.6 0.4021 0.11 8.0
73 1.030 1.9190 2.54 −0.8 0.4775 0.28 7.6 1.8769 2.80 −0.9 0.4045 0.22 6.7
74 1.044 1.9552 2.52 −1.5 0.4823 0.38 6.4 1.9168 2.78 −2.2 0.4083 0.30 5.1
75 1.058 1.9911 2.49 −2.8 0.4884 0.46 4.8 1.9563 2.74 −3.6 0.4132 0.36 3.7
76 1.072 2.0265 2.44 −4.3 0.4955 0.52 3.0 1.9951 2.68 −4.9 0.4188 0.41 2.3
77 1.087 2.0610 2.37 −5.7 0.5032 0.55 1.0 2.0329 2.60 −6.1 0.4249 0.43 0.7
78 1.101 2.0943 2.28 −6.8 0.5112 0.55 −1.1 2.0694 2.50 −7.0 0.4311 0.43 −1.2
79 1.115 2.1262 2.18 −7.4 0.5189 0.52 −2.8 2.1044 2.40 −7.7 0.4371 0.40 −2.9
80 1.130 2.1565 2.07 −7.7 0.5260 0.47 −4.1 2.1379 2.28 −8.0 0.4424 0.34 −4.2
81 1.144 2.1853 1.96 −7.7 0.5323 0.40 −5.0 2.1697 2.17 −8.0 0.4469 0.28 −5.0
82 1.158 2.2124 1.85 −7.6 0.5375 0.32 −6.0 2.1999 2.05 −8.1 0.4504 0.20 −5.7
83 1.173 2.2381 1.74 −7.4 0.5416 0.23 −6.9 2.2284 1.94 −8.0 0.4527 0.12 −6.1
84 1.187 2.2622 1.63 −7.2 0.5442 0.13 −7.9 2.2553 1.82 −7.7 0.4537 0.03 −6.2
85 1.201 2.2848 1.53 −6.9 0.5452 0.01 −8.7 2.2805 1.72 −6.9 0.4534 −0.06 −5.9
86 1.215 2.3060 1.44 −6.3 0.5443 −0.12 −8.9 2.3043 1.63 −5.5 0.4519 −0.14 −5.1
87 1.230 2.3259 1.35 −5.4 0.5416 −0.25 −8.1 2.3270 1.56 −3.9 0.4493 −0.21 −4.0
88 1.244 2.3446 1.28 −4.0 0.5372 −0.35 −6.2 2.3489 1.51 −2.5 0.4459 −0.26 −2.4
89 1.258 2.3625 1.24 −2.0 0.5315 −0.43 −3.6 2.3703 1.49 −1.5 0.4420 −0.28 −0.8
90 1.273 2.3800 1.22 0.3 0.5250 −0.46 −0.7 2.3914 1.47 −0.9 0.4380 −0.28 0.7
91 1.287 2.3975 1.25 2.9 0.5184 −0.45 2.2 2.4124 1.46 −0.3 0.4341 −0.26 1.6
92 1.301 2.4156 1.31 5.4 0.5122 −0.39 4.7 2.4333 1.46 0.7 0.4305 −0.23 1.7
93 1.316 2.4349 1.40 6.9 0.5071 −0.31 6.2 2.4542 1.48 1.7 0.4274 −0.21 1.6
94 1.330 2.4556 1.51 7.1 0.5033 −0.22 6.7 2.4756 1.51 2.5 0.4246 −0.19 1.7
ISTUDY
95 1.344 2.4779 1.60 5.8 0.5009 −0.12 6.4 2.4975 1.55 2.6 0.4220 −0.16 2.3
96 1.358 2.5014 1.67 3.7 0.4998 −0.03 5.4 2.5200 1.59 2.1 0.4200 −0.12 3.3
HCR 97 1.373 2.5257 1.71 1.3 0.4999 0.03 4.3 2.5429 1.61 1.2 0.4186 −0.07 4.1
98 1.387 2.5502 1.71 −0.7 0.5008 0.09 3.4 2.5661 1.62 −0.1 0.4180 0.00 4.4
99 1.401 2.5746 1.69 −2.4 0.5025 0.13 2.7 2.5893 1.61 −1.9 0.4185 0.06 3.9
100 1.416 2.5984 1.64 −4.3 0.5047 0.17 1.6 2.6121 1.57 −4.4 0.4197 0.11 2.2
101 1.430 2.6214 1.56 −6.8 0.5072 0.18 0.0 2.6341 1.48 −7.5 0.4215 0.12 0.0
102 1.444 2.6431 1.45 −9.5 0.5098 0.17 −1.5 2.6545 1.35 −10.6 0.4232 0.11 −1.7
103 1.459 2.6628 1.29 −11.7 0.5120 0.13 −2.5 2.6727 1.18 −12.7 0.4246 0.07 −2.5
104 1.473 2.6800 1.11 −12.7 0.5136 0.10 −2.6 2.6883 0.99 −13.3 0.4253 0.04 −2.2
105 1.487 2.6945 0.93 −12.3 0.5147 0.06 −2.2 2.7010 0.80 −12.6 0.4256 0.01 −1.6
106 1.501 2.7065 0.76 −11.4 0.5153 0.03 −1.7 2.7111 0.63 −11.4 0.4256 −0.01 −1.0
ISTUDY
TABLE A.2c Filtered Marker Kinematics – Lateral Malleolus (Ankle) and Heel
Right Ankle Right Heel
TOR 1 0.000 0.0939 2.24 19.1 0.2143 0.80 −6.3 0.0300 2.39 17.2 0.2360 1.32 −15.9
2 0.014 0.1279 2.50 16.4 0.2251 0.68 −10.0 0.0659 2.59 12.0 0.2531 1.03 −22.6
3 0.029 0.1653 2.71 13.5 0.2337 0.51 −13.2 0.1042 2.73 8.0 0.2655 0.67 −26.3
4 0.043 0.2054 2.88 10.8 0.2396 0.30 −15.2 0.1440 2.82 5.7 0.2723 0.28 −27.2
5 0.057 0.2477 3.02 8.7 0.2423 0.07 −16.0 0.1849 2.89 4.8 0.2734 −0.11 −26.1
6 0.072 0.2917 3.13 7.4 0.2417 −0.16 −15.5 0.2267 2.96 5.4 0.2692 −0.47 −23.7
7 0.086 0.3372 3.23 6.7 0.2379 −0.37 −14.1 0.2696 3.05 7.0 0.2600 −0.79 −20.4
8 0.100 0.3841 3.32 6.4 0.2311 −0.56 −12.1 0.3138 3.16 9.0 0.2466 −1.05 −16.4
9 0.114 0.4322 3.41 6.5 0.2219 −0.72 −9.7 0.3599 3.30 10.5 0.2299 −1.26 −11.9
10 0.129 0.4817 3.51 6.7 0.2106 −0.84 −7.1 0.4083 3.46 11.3 0.2107 −1.39 −7.3
11 0.143 0.5326 3.60 6.8 0.1979 −0.92 −4.0 0.4590 3.63 11.2 0.1900 −1.47 −3.0
12 0.157 0.5848 3.70 6.4 0.1844 −0.95 −0.5 0.5120 3.78 10.5 0.1687 −1.48 1.0
13 0.172 0.6384 3.79 5.4 0.1707 −0.93 3.0 0.5671 3.93 9.3 0.1477 −1.44 4.5
14 0.186 0.6931 3.85 3.7 0.1577 −0.86 6.2 0.6242 4.05 7.5 0.1276 −1.35 7.3
15 0.200 0.7486 3.89 1.3 0.1460 −0.76 8.6 0.6829 4.14 5.2 0.1090 −1.23 9.7
16 0.215 0.8045 3.89 −1.6 0.1360 −0.62 10.1 0.7427 4.20 2.1 0.0925 −1.07 11.6
17 0.229 0.8600 3.85 −5.1 0.1282 −0.47 10.7 0.8029 4.20 −1.7 0.0783 −0.90 12.9
18 0.243 0.9145 3.75 −8.8 0.1226 −0.31 10.3 0.8628 4.15 −6.2 0.0668 −0.70 13.3
19 0.257 0.9672 3.60 −12.5 0.1192 −0.17 8.9 0.9215 4.02 −11.4 0.0582 −0.51 12.6
20 0.272 1.0173 3.39 −16.2 0.1177 −0.06 6.7 0.9779 3.82 −17.1 0.0521 −0.34 10.8
21 0.286 1.0641 3.13 −20.0 0.1175 0.02 3.9 1.0309 3.53 −23.2 0.0483 −0.21 8.1
22 0.300 1.1068 2.82 −23.6 0.1182 0.05 0.7 1.0790 3.16 −28.9 0.0462 −0.11 4.6
23 0.315 1.1447 2.46 −26.3 0.1190 0.04 −2.5 1.1212 2.71 −33.1 0.0451 −0.07 1.1
24 0.329 1.1771 2.07 −27.4 0.1193 −0.02 −5.1 1.1565 2.21 −35.2 0.0441 −0.08 −1.1
25 0.343 1.2038 1.67 −26.3 0.1185 −0.11 −6.5 1.1844 1.70 −34.7 0.0428 −0.11 −1.2
26 0.357 1.2249 1.31 −23.1 0.1162 −0.21 −6.1 1.2051 1.22 −31.6 0.0411 −0.12 0.4
27 0.372 1.2413 1.01 −18.3 0.1126 −0.28 −3.7 1.2192 0.80 −26.1 0.0395 −0.09 2.4
HCR 28 0.386 1.2539 0.79 −13.3 0.1081 −0.31 −0.2 1.2280 0.47 −19.3 0.0384 −0.05 3.5
29 0.400 1.2639 0.63 −9.2 0.1037 −0.29 3.0 1.2327 0.25 −12.5 0.0381 0.01 3.3
30 0.415 1.2720 0.53 −6.7 0.0999 −0.23 4.6 1.2350 0.11 −6.8 0.0385 0.05 2.1
ISTUDY
31 0.429 1.2789 0.44 −5.5 0.0972 −0.16 4.7 1.2360 0.05 −2.7 0.0394 0.07 0.9
32 0.443 1.2846 0.37 −5.1 0.0955 −0.09 3.7 1.2365 0.04 −0.4 0.0404 0.07 0.1
33 0.458 1.2894 0.30 −4.8 0.0945 −0.05 2.3 1.2370 0.04 0.5 0.0414 0.07 −0.3
34 0.472 1.2931 0.23 −4.3 0.0940 −0.03 1.2 1.2377 0.05 0.8 0.0424 0.06 −0.4
35 0.486 1.2960 0.18 −3.3 0.0937 −0.02 0.6 1.2384 0.06 0.9 0.0432 0.06 −0.3
36 0.500 1.2982 0.14 −2.4 0.0935 −0.01 0.5 1.2394 0.08 0.9 0.0440 0.05 −0.5
37 0.515 1.2999 0.11 −1.9 0.0934 0.00 0.5 1.2406 0.09 0.2 0.0448 0.05 −0.8
38 0.529 1.3012 0.08 −2.1 0.0935 0.01 0.4 1.2419 0.08 −1.1 0.0453 0.03 −1.0
39 0.543 1.3022 0.05 −2.2 0.0936 0.01 0.3 1.2429 0.06 −2.2 0.0457 0.02 −0.9
40 0.558 1.3026 0.02 −2.0 0.0938 0.01 0.5 1.2435 0.02 −2.5 0.0458 0.00 −0.6
41 0.572 1.3026 −0.01 −1.3 0.0940 0.02 1.0 1.2435 −0.01 −1.8 0.0458 0.00 0.0
42 0.586 1.3023 −0.02 −0.5 0.0945 0.04 1.3 1.2431 −0.03 −0.7 0.0458 0.00 0.6
43 0.601 1.3020 −0.02 0.3 0.0952 0.06 1.2 1.2425 −0.04 0.3 0.0459 0.02 1.1
44 0.615 1.3017 −0.01 0.9 0.0962 0.08 0.6 1.2421 −0.02 1.1 0.0463 0.04 1.0
45 0.629 1.3016 0.00 1.5 0.0974 0.08 −0.2 1.2418 0.00 1.6 0.0469 0.05 0.5
46 0.643 1.3018 0.03 1.7 0.0985 0.07 −0.9 1.2420 0.02 1.8 0.0477 0.05 −0.1
47 0.658 1.3024 0.05 1.6 0.0994 0.05 −1.1 1.2425 0.05 1.7 0.0484 0.04 −0.5
48 0.672 1.3034 0.07 1.3 0.1000 0.04 −0.9 1.2433 0.07 1.6 0.0489 0.03 −0.5
49 0.686 1.3046 0.09 0.9 0.1005 0.03 −0.3 1.2445 0.09 1.3 0.0494 0.03 0.1
50 0.701 1.3059 0.10 0.5 0.1009 0.03 0.3 1.2460 0.11 0.9 0.0498 0.04 1.2
51 0.715 1.3074 0.11 0.1 0.1014 0.04 0.8 1.2477 0.12 0.2 0.0504 0.06 2.3
52 0.729 1.3090 0.10 −0.4 0.1020 0.05 1.4 1.2494 0.12 −0.6 0.0516 0.10 3.1
53 0.744 1.3104 0.10 −0.5 0.1029 0.08 1.9 1.2510 0.10 −0.9 0.0534 0.15 3.4
54 0.758 1.3117 0.09 −0.2 0.1042 0.11 2.6 1.2523 0.09 −0.6 0.0559 0.20 3.5
55 0.772 1.3129 0.09 0.3 0.1061 0.15 3.2 1.2536 0.08 −0.3 0.0591 0.25 3.5
56 0.786 1.3142 0.10 0.8 0.1086 0.20 3.6 1.2547 0.08 −0.1 0.0631 0.30 3.5
57 0.801 1.3157 0.11 1.4 0.1118 0.25 3.6 1.2559 0.08 0.3 0.0677 0.35 3.8
58 0.815 1.3174 0.14 2.7 0.1158 0.30 3.2 1.2571 0.09 1.5 0.0731 0.41 4.6
59 0.829 1.3196 0.19 4.7 0.1205 0.35 3.2 1.2585 0.12 3.5 0.0794 0.48 5.9
60 0.844 1.3227 0.27 7.2 0.1257 0.39 3.5 1.2606 0.19 6.1 0.0869 0.58 7.4
61 0.858 1.3274 0.39 9.8 0.1317 0.45 4.0 1.2640 0.30 9.1 0.0960 0.69 8.5
62 0.872 1.3340 0.55 12.2 0.1385 0.51 4.4 1.2692 0.45 12.3 0.1068 0.82 9.3
63 0.887 1.3432 0.74 14.0 0.1463 0.57 4.5 1.2769 0.65 15.4 0.1195 0.96 9.9
64 0.901 1.3552 0.95 15.0 0.1549 0.64 4.4 1.2878 0.89 17.8 0.1343 1.11 10.0
(continued )
ISTUDY
TABLE A.2c (Continued)
Right Ankle Right Heel
65 0.915 1.3704 1.17 15.3 0.1645 0.70 3.6 1.3024 1.16 19.1 0.1511 1.25 9.1
66 0.929 1.3887 1.39 15.2 0.1749 0.74 2.2 1.3210 1.44 19.2 0.1700 1.37 6.1
67 0.944 1.4101 1.61 15.1 0.1857 0.76 0.6 1.3436 1.71 18.0 0.1902 1.42 0.8
68 0.958 1.4346 1.82 14.9 0.1967 0.76 −1.1 1.3699 1.95 15.8 0.2106 1.39 −5.7
69 0.972 1.4622 2.03 14.4 0.2074 0.73 −3.0 1.3995 2.16 12.9 0.2299 1.26 −11.7
TOR 70 0.987 1.4928 2.23 13.4 0.2175 0.67 −5.3 1.4317 2.32 9.9 0.2467 1.05 −16.4
71 1.001 1.5261 2.41 12.2 0.2266 0.58 −8.2 1.4659 2.44 7.8 0.2600 0.79 −19.6
72 1.015 1.5619 2.58 11.3 0.2340 0.44 −11.3 1.5016 2.55 7.3 0.2693 0.49 −21.8
73 1.030 1.5999 2.74 10.7 0.2391 0.25 −13.9 1.5387 2.65 7.8 0.2741 0.17 −22.9
74 1.044 1.6401 2.89 10.2 0.2413 0.04 −15.2 1.5774 2.77 8.5 0.2741 −0.16 −22.6
75 1.058 1.6824 3.03 9.7 0.2403 −0.18 −15.1 1.6178 2.89 9.0 0.2694 −0.48 −21.1
76 1.072 1.7267 3.16 9.3 0.2361 −0.39 −13.8 1.6601 3.03 9.4 0.2604 −0.76 −18.8
77 1.087 1.7729 3.29 8.8 0.2291 −0.58 −11.9 1.7044 3.16 9.8 0.2475 −1.02 −16.2
78 1.101 1.8209 3.42 8.2 0.2197 −0.73 −9.7 1.7506 3.31 10.3 0.2314 −1.23 −13.3
79 1.115 1.8706 3.53 7.5 0.2082 −0.85 −7.3 1.7990 3.46 10.8 0.2124 −1.40 −9.9
80 1.130 1.9218 3.63 6.9 0.1953 −0.94 −4.5 1.8495 3.62 11.2 0.1914 −1.51 −5.8
81 1.144 1.9745 3.72 6.2 0.1814 −0.98 −1.5 1.9024 3.78 11.3 0.1692 −1.56 −1.4
82 1.158 2.0284 3.81 5.5 0.1672 −0.98 1.7 1.9576 3.94 10.8 0.1467 −1.55 3.0
83 1.173 2.0834 3.88 4.6 0.1533 −0.93 5.0 2.0150 4.09 9.5 0.1249 −1.48 6.9
84 1.187 2.1394 3.94 3.3 0.1405 −0.84 8.1 2.0745 4.21 7.4 0.1044 −1.35 10.2
85 1.201 2.1961 3.98 1.3 0.1293 −0.70 10.6 2.1354 4.30 4.5 0.0862 −1.19 12.8
86 1.215 2.2531 3.98 −1.7 0.1204 −0.54 11.8 2.1974 4.34 0.6 0.0705 −0.99 14.5
87 1.230 2.3098 3.93 −5.8 0.1140 −0.36 11.6 2.2595 4.32 −4.2 0.0579 −0.77 15.0
88 1.244 2.3654 3.81 −10.9 0.1100 −0.21 10.2 2.3208 4.22 −10.1 0.0485 −0.56 14.3
89 1.258 2.4188 3.61 −16.6 0.1081 −0.07 7.9 2.3801 4.03 −16.9 0.0420 −0.36 12.4
90 1.273 2.4688 3.33 −22.3 0.1079 0.02 4.8 2.4360 3.73 −24.2 0.0381 −0.20 9.5
91 1.287 2.5142 2.98 −26.8 0.1087 0.07 1.1 2.4869 3.34 −30.9 0.0362 −0.09 6.0
92 1.301 2.5539 2.57 −29.4 0.1098 0.05 −2.6 2.5314 2.85 −35.8 0.0356 −0.03 2.6
93 1.316 2.5876 2.14 −29.4 0.1103 −0.01 −5.3 2.5684 2.31 −38.0 0.0354 −0.01 0.2
94 1.330 2.6151 1.73 −27.1 0.1095 −0.10 −6.1 2.5975 1.76 −37.1 0.0352 −0.02 −0.5
ISTUDY
95 1.344 2.6369 1.36 −22.7 0.1074 −0.18 −4.5 2.6188 1.25 −33.1 0.0347 −0.03 0.3
96 1.358 2.6541 1.08 −17.4 0.1043 −0.23 −1.6 2.6333 0.82 −26.6 0.0343 −0.01 1.7
HCR 97 1.373 2.6677 0.87 −12.5 0.1009 −0.23 1.2 2.6422 0.49 −19.1 0.0343 0.02 2.4
98 1.387 2.6789 0.72 −9.3 0.0978 −0.19 2.9 2.6473 0.27 −12.1 0.0349 0.05 2.0
99 1.401 2.6882 0.60 −8.0 0.0954 −0.15 3.1 2.6500 0.14 −6.9 0.0359 0.08 0.8
100 1.416 2.6960 0.49 −7.8 0.0936 −0.10 2.5 2.6514 0.07 −3.7 0.0371 0.08 −0.6
101 1.430 2.7023 0.38 −7.8 0.0924 −0.08 1.5 2.6521 0.04 −2.2 0.0381 0.06 −1.7
102 1.444 2.7068 0.27 −7.7 0.0915 −0.06 0.7 2.6524 0.01 −1.5 0.0388 0.03 −2.3
103 1.459 2.7098 0.16 −7.0 0.0907 −0.05 0.4 2.6525 0.00 −1.1 0.0390 0.00 −2.2
104 1.473 2.7114 0.07 −5.5 0.0899 −0.05 0.5 2.6523 −0.02 −0.8 0.0387 −0.03 −1.5
105 1.487 2.7117 0.00 −3.4 0.0893 −0.04 0.7 2.6519 −0.03 −0.4 0.0381 −0.05 −0.6
106 1.501 2.7114 −0.03 −1.2 0.0888 −0.03 0.8 2.6515 −0.03 −0.1 0.0373 −0.05 0.2
ISTUDY
TABLE A.2d Filtered Marker Kinematics – Fifth Metatarsal and Toe
Right Metatarsal Right Toe
TOR 1 0.000 0.0845 1.63 33.7 0.0926 0.83 −3.5 0.1283 1.34 38.1 0.0464 0.35 5.6
2 0.014 0.1114 2.12 33.5 0.1042 0.74 −9.6 0.1515 1.92 40.8 0.0521 0.40 1.3
3 0.029 0.1452 2.59 30.8 0.1137 0.56 −14.4 0.1833 2.51 39.0 0.0580 0.38 −4.1
4 0.043 0.1854 3.00 26.7 0.1201 0.33 −17.1 0.2232 3.04 34.1 0.0630 0.29 −8.4
5 0.057 0.2311 3.35 22.4 0.1230 0.07 −17.4 0.2702 3.48 28.1 0.0662 0.14 −10.3
6 0.072 0.2813 3.64 18.4 0.1222 −0.17 −15.9 0.3229 3.84 22.3 0.0671 −0.01 −9.8
7 0.086 0.3353 3.88 15.0 0.1180 −0.38 −13.0 0.3801 4.12 17.1 0.0659 −0.14 −7.7
8 0.100 0.3923 4.07 12.0 0.1112 −0.55 −9.2 0.4407 4.33 12.7 0.0631 −0.23 −4.6
9 0.114 0.4517 4.22 9.4 0.1024 −0.65 −4.8 0.5039 4.48 8.9 0.0594 −0.27 −1.2
10 0.129 0.5130 4.34 7.1 0.0927 −0.68 −0.5 0.5689 4.59 5.6 0.0554 −0.26 2.1
11 0.143 0.5758 4.43 5.0 0.0828 −0.66 3.2 0.6351 4.64 2.8 0.0518 −0.21 5.2
12 0.157 0.6396 4.48 3.0 0.0737 −0.59 6.5 0.7018 4.67 0.3 0.0493 −0.12 7.8
13 0.172 0.7040 4.51 0.9 0.0659 −0.48 9.2 0.7685 4.65 −2.1 0.0485 0.01 10.1
14 0.186 0.7686 4.51 −1.3 0.0601 −0.33 11.1 0.8349 4.61 −4.5 0.0497 0.17 11.7
15 0.200 0.8330 4.47 −3.6 0.0565 −0.16 12.3 0.9003 4.52 −7.1 0.0534 0.35 12.6
16 0.215 0.8966 4.41 −5.8 0.0555 0.02 12.7 0.9642 4.40 −9.9 0.0596 0.53 12.8
17 0.229 0.9590 4.31 −8.5 0.0572 0.20 12.5 1.0262 4.24 −13.2 0.0686 0.71 11.9
18 0.243 1.0197 4.16 −12.1 0.0614 0.38 11.2 1.0855 4.03 −16.8 0.0801 0.87 9.5
19 0.257 1.0780 3.96 −16.6 0.0680 0.53 8.7 1.1414 3.76 −20.2 0.0936 0.99 5.7
20 0.272 1.1330 3.69 −21.4 0.0765 0.63 4.5 1.1930 3.45 −23.1 0.1083 1.04 0.6
21 0.286 1.1835 3.35 −25.7 0.0859 0.66 −0.9 1.2400 3.10 −25.2 0.1232 1.00 −5.5
22 0.300 1.2288 2.95 −28.9 0.0952 0.60 −7.2 1.2817 2.73 −26.4 0.1370 0.88 −12.1
23 0.315 1.2680 2.52 −30.8 0.1031 0.45 −13.2 1.3180 2.35 −26.8 0.1483 0.66 −18.5
24 0.329 1.3009 2.07 −31.0 0.1081 0.22 −17.6 1.3488 1.96 −26.1 0.1558 0.35 −23.5
25 0.343 1.3274 1.63 −29.5 0.1094 −0.05 −19.2 1.3741 1.60 −23.9 0.1583 −0.01 −25.5
26 0.357 1.3477 1.23 −25.9 0.1066 −0.33 −17.7 1.3945 1.28 −20.3 0.1554 −0.38 −24.1
27 0.372 1.3626 0.89 −21.0 0.1001 −0.56 −13.5 1.4107 1.02 −15.8 0.1474 −0.70 −19.4
HCR 28 0.386 1.3732 0.63 −15.6 0.0907 −0.71 −7.6 1.4237 0.83 −11.9 0.1353 −0.94 −12.6
29 0.400 1.3807 0.45 −10.9 0.0797 −0.78 −1.4 1.4343 0.68 −9.2 0.1206 −1.06 −4.9
30 0.415 1.3860 0.32 −7.3 0.0685 −0.75 4.0 1.4432 0.56 −7.7 0.1050 −1.08 2.4
ISTUDY
31 0.429 1.3899 0.24 −5.1 0.0582 −0.66 7.9 1.4504 0.46 −6.9 0.0898 −0.99 8.2
32 0.443 1.3928 0.18 −3.7 0.0496 −0.53 9.9 1.4563 0.37 −6.2 0.0765 −0.84 11.9
33 0.458 1.3949 0.13 −2.8 0.0431 −0.38 10.2 1.4609 0.28 −5.6 0.0658 −0.65 13.2
34 0.472 1.3965 0.10 −1.9 0.0387 −0.24 9.0 1.4644 0.21 −4.8 0.0578 −0.46 12.3
35 0.486 1.3977 0.08 −1.0 0.0363 −0.12 6.8 1.4668 0.15 −3.7 0.0525 −0.30 10.3
36 0.500 1.3987 0.07 −0.4 0.0353 −0.04 4.2 1.4685 0.10 −2.5 0.0492 −0.17 7.7
37 0.515 1.3997 0.07 −0.5 0.0351 0.00 2.0 1.4697 0.07 −1.8 0.0476 −0.08 5.0
38 0.529 1.4006 0.06 −1.2 0.0353 0.01 0.5 1.4706 0.05 −1.7 0.0469 −0.03 2.4
39 0.543 1.4013 0.03 −1.9 0.0355 0.01 −0.1 1.4712 0.02 −1.7 0.0467 −0.01 0.5
40 0.558 1.4016 0.00 −1.8 0.0357 0.01 −0.2 1.4713 0.00 −1.4 0.0466 −0.01 −0.5
41 0.572 1.4014 −0.02 −1.2 0.0359 0.01 −0.1 1.4712 −0.02 −0.9 0.0463 −0.02 −0.7
42 0.586 1.4010 −0.03 −0.3 0.0360 0.01 −0.2 1.4709 −0.02 −0.4 0.0459 −0.03 −0.5
43 0.601 1.4005 −0.03 0.3 0.0361 0.00 −0.5 1.4705 −0.03 0.1 0.0453 −0.04 −0.1
44 0.615 1.4001 −0.02 0.6 0.0361 −0.01 −0.9 1.4701 −0.02 0.5 0.0448 −0.04 0.0
45 0.629 1.3998 −0.01 0.6 0.0359 −0.02 −0.8 1.4699 −0.01 0.7 0.0443 −0.04 −0.1
46 0.643 1.3997 0.00 0.5 0.0355 −0.03 −0.3 1.4698 0.00 0.7 0.0437 −0.04 −0.3
47 0.658 1.3997 0.00 0.3 0.0351 −0.03 0.2 1.4698 0.01 0.7 0.0431 −0.05 −0.3
48 0.672 1.3997 0.00 0.5 0.0346 −0.02 0.6 1.4700 0.02 0.7 0.0424 −0.05 −0.2
49 0.686 1.3998 0.01 0.9 0.0344 −0.01 0.6 1.4703 0.03 0.6 0.0417 −0.05 0.0
50 0.701 1.4001 0.03 1.2 0.0343 −0.01 0.2 1.4708 0.04 0.4 0.0410 −0.05 0.3
51 0.715 1.4006 0.05 1.0 0.0342 −0.01 −0.2 1.4714 0.04 0.0 0.0403 −0.04 0.5
52 0.729 1.4014 0.06 0.5 0.0340 −0.01 −0.1 1.4719 0.03 −0.5 0.0397 −0.03 0.6
53 0.744 1.4023 0.06 −0.1 0.0338 −0.01 0.6 1.4724 0.02 −0.8 0.0393 −0.03 0.6
54 0.758 1.4031 0.05 −0.6 0.0337 0.00 1.3 1.4726 0.01 −0.9 0.0390 −0.02 0.8
55 0.772 1.4038 0.04 −1.1 0.0339 0.03 1.6 1.4727 0.00 −1.0 0.0388 0.00 0.8
56 0.786 1.4043 0.02 −1.6 0.0345 0.05 1.1 1.4725 −0.02 −1.3 0.0389 0.01 0.7
57 0.801 1.4045 0.00 −1.9 0.0353 0.06 0.3 1.4721 −0.04 −1.4 0.0390 0.01 0.2
58 0.815 1.4042 −0.03 −1.5 0.0361 0.06 −0.4 1.4714 −0.06 −1.0 0.0393 0.01 −0.3
59 0.829 1.4035 −0.05 −0.4 0.0369 0.05 −0.5 1.4705 −0.07 −0.1 0.0394 0.00 −1.0
60 0.844 1.4028 −0.04 1.2 0.0374 0.04 0.2 1.4695 −0.06 1.0 0.0394 −0.02 −1.7
61 0.858 1.4023 −0.01 3.0 0.0380 0.05 1.8 1.4687 −0.04 2.1 0.0390 −0.04 −2.1
62 0.872 1.4024 0.04 4.6 0.0389 0.09 4.0 1.4684 0.00 2.8 0.0381 −0.08 −1.8
63 0.887 1.4035 0.12 5.8 0.0406 0.17 6.4 1.4687 0.04 3.2 0.0368 −0.10 −0.8
64 0.901 1.4058 0.21 7.2 0.0437 0.27 8.3 1.4696 0.09 3.5 0.0353 −0.10 1.0
(continued )
ISTUDY
TABLE A.2d (Continued)
Right Metatarsal Right Toe
65 0.915 1.4094 0.32 9.2 0.0485 0.40 9.3 1.4712 0.14 4.6 0.0340 −0.07 3.0
66 0.929 1.4149 0.47 12.4 0.0552 0.54 9.0 1.4737 0.22 7.3 0.0333 −0.01 5.0
67 0.944 1.4229 0.68 16.9 0.0639 0.66 7.3 1.4775 0.35 12.7 0.0337 0.07 6.3
68 0.958 1.4343 0.95 22.1 0.0741 0.75 4.1 1.4837 0.58 20.7 0.0354 0.17 6.4
69 0.972 1.4502 1.31 26.8 0.0852 0.78 −0.2 1.4941 0.94 29.5 0.0385 0.26 5.4
TOR 70 0.987 1.4717 1.72 29.7 0.0963 0.74 −5.4 1.5107 1.43 36.3 0.0428 0.32 3.4
71 1.001 1.4994 2.16 30.2 0.1063 0.62 −10.6 1.5349 1.98 39.0 0.0478 0.35 0.3
72 1.015 1.5335 2.59 28.5 0.1142 0.44 −14.7 1.5673 2.54 37.5 0.0530 0.33 −3.4
73 1.030 1.5734 2.97 25.3 0.1188 0.20 −16.9 1.6075 3.05 33.2 0.0574 0.26 −7.0
74 1.044 1.6185 3.31 21.7 0.1200 −0.05 −16.6 1.6546 3.49 27.9 0.0603 0.13 −9.3
75 1.058 1.6681 3.59 18.2 0.1175 −0.27 −14.3 1.7074 3.85 22.5 0.0612 −0.01 −9.9
76 1.072 1.7213 3.83 15.2 0.1121 −0.45 −10.8 1.7647 4.13 17.5 0.0601 −0.15 −8.5
77 1.087 1.7776 4.03 12.7 0.1046 −0.58 −7.3 1.8256 4.35 13.2 0.0570 −0.25 −5.7
78 1.101 1.8365 4.19 10.4 0.0955 −0.66 −4.3 1.8892 4.51 9.7 0.0528 −0.31 −2.1
79 1.115 1.8975 4.33 8.3 0.0856 −0.70 −1.4 1.9547 4.63 6.7 0.0482 −0.31 1.5
80 1.130 1.9603 4.43 6.4 0.0754 −0.70 1.7 2.0215 4.70 4.0 0.0439 −0.27 4.7
81 1.144 2.0243 4.51 4.7 0.0655 −0.66 4.9 2.0892 4.74 1.0 0.0406 −0.18 7.5
82 1.158 2.0892 4.57 3.1 0.0566 −0.56 8.1 2.1571 4.73 −1.9 0.0388 −0.05 9.9
83 1.173 2.1548 4.60 1.5 0.0493 −0.43 10.9 2.2246 4.69 −4.5 0.0391 0.11 12.0
84 1.187 2.2208 4.61 −0.6 0.0444 −0.25 13.2 2.2912 4.61 −6.5 0.0418 0.29 13.4
85 1.201 2.2866 4.58 −3.4 0.0421 −0.05 14.7 2.3564 4.50 −8.4 0.0474 0.49 14.0
86 1.215 2.3518 4.51 −7.0 0.0430 0.17 15.1 2.4199 4.37 −10.7 0.0558 0.69 13.7
87 1.230 2.4155 4.38 −11.3 0.0469 0.38 14.2 2.4813 4.20 −13.9 0.0671 0.88 12.1
88 1.244 2.4771 4.19 −16.3 0.0539 0.57 11.6 2.5399 3.97 −18.0 0.0811 1.04 8.4
89 1.258 2.5353 3.91 −21.8 0.0634 0.71 6.8 2.5948 3.68 −22.4 0.0968 1.12 2.6
90 1.273 2.5890 3.56 −27.2 0.0743 0.77 0.1 2.6451 3.33 −26.3 0.1132 1.11 −4.7
91 1.287 2.6371 3.14 −31.4 0.0853 0.72 −7.3 2.6900 2.93 −28.8 0.1286 0.99 −12.6
92 1.301 2.6787 2.66 −33.6 0.0948 0.56 −14.1 2.7289 2.51 −29.5 0.1415 0.75 −20.0
93 1.316 2.7133 2.18 −33.4 0.1013 0.31 −18.8 2.7617 2.09 −28.2 0.1501 0.42 −25.4
94 1.330 2.7410 1.71 −30.8 0.1038 0.02 −20.6 2.7886 1.70 −25.4 0.1534 0.02 −27.6
ISTUDY
95 1.344 2.7622 1.30 −26.1 0.1019 −0.28 −19.0 2.8103 1.36 −21.4 0.1508 −0.37 −25.8
96 1.358 2.7781 0.96 −20.4 0.0959 −0.52 −14.5 2.8275 1.09 −16.8 0.1427 −0.72 −20.6
HCR 97 1.373 2.7898 0.71 −15.0 0.0869 −0.69 −8.3 2.8413 0.88 −12.6 0.1304 −0.96 −12.9
98 1.387 2.7985 0.53 −11.0 0.0762 −0.76 −1.9 2.8527 0.73 −10.0 0.1153 −1.08 −4.2
99 1.401 2.8050 0.40 −8.5 0.0651 −0.75 3.5 2.8621 0.60 −9.0 0.0994 −1.08 3.6
100 1.416 2.8099 0.29 −7.2 0.0548 −0.66 7.1 2.8697 0.47 −9.0 0.0843 −0.98 9.3
101 1.430 2.8134 0.20 −6.2 0.0461 −0.54 8.6 2.8755 0.34 −8.9 0.0714 −0.81 12.4
102 1.444 2.8155 0.11 −5.4 0.0393 −0.42 8.5 2.8794 0.21 −8.0 0.0610 −0.63 12.9
103 1.459 2.8166 0.04 −4.4 0.0342 −0.30 7.4 2.8816 0.11 −6.3 0.0535 −0.44 11.6
104 1.473 2.8167 −0.01 −2.9 0.0306 −0.21 6.1 2.8825 0.03 −4.1 0.0483 −0.29 9.2
105 1.487 2.8162 −0.04 −1.1 0.0283 −0.13 4.9 2.8825 −0.01 −1.8 0.0451 −0.18 6.5
106 1.501 2.8155 −0.04 0.3 0.0270 −0.07 3.5 2.8822 −0.02 −0.1 0.0431 −0.11 4.1
ISTUDY
TABLE A.3a Linear and Angular Kinematics – Foot
Frame Time Theta Omega Alpha Cofm-X Vel-X Acc-X Cofm-Y Vel-Y Acc-Y
s Deg R/s R/s/s m m/s m/s/s m m/s m/s/s
TOR 1 0.000 85.6 −5.01 118.27 0.089 1.937 26.39 0.153 0.814 −4.92
2 0.014 82.2 −3.11 138.72 0.120 2.310 24.92 0.165 0.707 −9.81
3 0.029 80.5 −1.04 142.01 0.155 2.650 22.11 0.174 0.533 −13.77
4 0.043 80.5 0.96 132.71 0.195 2.942 18.72 0.180 0.313 −16.13
5 0.057 82.1 2.75 115.36 0.239 3.185 15.54 0.183 0.072 −16.70
6 0.072 85.0 4.25 93.06 0.287 3.386 12.91 0.182 −0.164 −15.72
7 0.086 89.1 5.41 68.58 0.336 3.555 10.83 0.178 −0.378 −13.58
8 0.100 93.9 6.22 44.92 0.388 3.696 9.20 0.171 −0.553 −10.64
9 0.114 99.2 6.70 25.05 0.442 3.818 7.94 0.162 −0.682 −7.27
10 0.129 104.9 6.93 10.88 0.497 3.923 6.90 0.152 −0.761 −3.80
11 0.143 110.6 7.01 2.33 0.554 4.015 5.88 0.140 −0.791 −0.37
12 0.157 116.3 7.00 −2.26 0.612 4.092 4.69 0.129 −0.771 2.99
13 0.172 122.1 6.94 −4.67 0.671 4.149 3.15 0.118 −0.705 6.11
14 0.186 127.7 6.87 −5.67 0.731 4.182 1.18 0.109 −0.597 8.65
15 0.200 133.3 6.78 −5.01 0.791 4.183 −1.12 0.101 −0.457 10.42
16 0.215 138.8 6.72 −3.24 0.851 4.150 −3.73 0.096 −0.299 11.39
17 0.229 144.3 6.69 −3.16 0.909 4.076 −6.80 0.093 −0.132 11.57
18 0.243 149.8 6.63 −8.11 0.967 3.955 −10.45 0.092 0.032 10.78
19 0.257 155.2 6.46 −19.17 1.023 3.778 −14.54 0.094 0.177 8.80
20 0.272 160.4 6.08 −34.96 1.075 3.539 −18.79 0.097 0.284 5.61
21 0.286 165.2 5.46 −53.64 1.124 3.240 −22.83 0.102 0.337 1.47
22 0.300 169.3 4.55 −73.48 1.168 2.886 −26.24 0.107 0.326 −3.24
23 0.315 172.6 3.36 −91.19 1.206 2.490 −28.53 0.111 0.245 −7.83
24 0.329 174.8 1.94 −102.29 1.239 2.070 −29.23 0.114 0.102 −11.35
25 0.343 175.8 0.43 −104.08 1.266 1.654 −27.90 0.114 −0.080 −12.86
26 0.357 175.5 −1.03 −97.50 1.286 1.272 −24.51 0.111 −0.266 −11.90
27 0.372 174.1 −2.36 −85.55 1.302 0.953 −19.64 0.106 −0.420 −8.61
HCR 28 0.386 171.7 −3.48 −68.80 1.314 0.710 −14.43 0.099 −0.512 −3.92
29 0.400 168.4 −4.32 −43.96 1.322 0.540 −10.02 0.092 −0.532 0.78
30 0.415 164.6 −4.74 −10.50 1.329 0.424 −7.03 0.084 −0.490 4.32
31 0.429 160.6 −4.63 24.51 1.334 0.339 −5.31 0.078 −0.409 6.26
ISTUDY
32 0.443 157.0 −4.04 52.05 1.339 0.272 −4.39 0.073 −0.311 6.77
33 0.458 154.0 −3.14 67.24 1.342 0.214 −3.77 0.069 −0.215 6.26
34 0.472 151.9 −2.11 68.18 1.345 0.164 −3.06 0.066 −0.132 5.12
35 0.486 150.6 −1.19 55.74 1.347 0.126 −2.15 0.065 −0.069 3.71
36 0.500 149.9 −0.52 35.86 1.348 0.103 −1.38 0.064 −0.026 2.36
37 0.515 149.7 −0.16 17.08 1.350 0.087 −1.21 0.064 −0.001 1.25
38 0.529 149.7 −0.03 4.56 1.351 0.068 −1.64 0.064 0.010 0.46
39 0.543 149.7 −0.03 −1.79 1.352 0.040 −2.06 0.065 0.012 0.10
40 0.558 149.6 −0.08 −4.87 1.352 0.009 −1.91 0.065 0.013 0.16
41 0.572 149.5 −0.17 −7.39 1.352 −0.015 −1.21 0.065 0.017 0.43
42 0.586 149.3 −0.29 −10.28 1.352 −0.026 −0.40 0.065 0.025 0.54
43 0.601 149.0 −0.46 −12.53 1.351 −0.026 0.26 0.066 0.033 0.31
44 0.615 148.6 −0.65 −12.14 1.351 −0.018 0.76 0.066 0.034 −0.11
45 0.629 148.0 −0.81 −8.07 1.351 −0.004 1.07 0.067 0.029 −0.47
46 0.643 147.2 −0.88 −1.88 1.351 0.012 1.10 0.067 0.021 −0.60
47 0.658 146.5 −0.86 3.55 1.351 0.027 0.96 0.067 0.012 −0.45
48 0.672 145.8 −0.78 6.31 1.352 0.039 0.86 0.067 0.008 −0.13
49 0.686 145.2 −0.68 5.50 1.352 0.052 0.88 0.067 0.009 0.13
50 0.701 144.7 −0.62 1.74 1.353 0.065 0.85 0.068 0.012 0.22
51 0.715 144.2 −0.63 −2.68 1.354 0.076 0.55 0.068 0.015 0.32
52 0.729 143.7 −0.70 −5.69 1.355 0.080 0.05 0.068 0.021 0.66
53 0.744 143.1 −0.80 −7.38 1.356 0.077 −0.32 0.068 0.034 1.25
54 0.758 142.4 −0.91 −10.45 1.357 0.071 −0.42 0.069 0.056 1.91
55 0.772 141.6 −1.10 −17.81 1.358 0.065 −0.41 0.070 0.088 2.38
56 0.786 140.6 −1.42 −28.50 1.359 0.059 −0.42 0.072 0.125 2.38
57 0.801 139.2 −1.91 −38.25 1.360 0.053 −0.24 0.074 0.156 1.90
58 0.815 137.4 −2.52 −45.01 1.361 0.053 0.57 0.076 0.179 1.39
59 0.829 135.1 −3.20 −49.94 1.362 0.069 2.14 0.079 0.196 1.33
60 0.844 132.2 −3.94 −52.91 1.363 0.114 4.22 0.082 0.217 1.88
61 0.858 128.6 −4.71 −52.39 1.365 0.190 6.40 0.085 0.250 2.88
62 0.872 124.5 −5.44 −49.61 1.368 0.297 8.37 0.089 0.300 4.16
63 0.887 119.7 −6.13 −47.15 1.373 0.429 9.90 0.093 0.369 5.46
64 0.901 114.4 −6.79 −43.01 1.380 0.580 11.08 0.099 0.456 6.34
65 0.915 108.6 −7.36 −31.69 1.390 0.746 12.24 0.106 0.550 6.44
(continued )
ISTUDY
66 0.929 102.4 −7.70 −10.02 1.402 0.930 13.82 0.115 0.640 5.61
67 0.944 96.0 −7.65 21.54 1.417 1.141 16.00 0.125 0.711 3.93
68 0.958 89.8 −7.08 60.42 1.434 1.388 18.51 0.135 0.752 1.50
69 0.972 84.4 −5.92 100.83 1.456 1.671 20.60 0.146 0.754 −1.62
TOR 70 0.987 80.1 −4.20 133.15 1.482 1.977 21.55 0.157 0.706 −5.36
71 1.001 77.5 −2.11 148.09 1.513 2.287 21.20 0.166 0.600 −9.39
72 1.015 76.7 0.04 142.55 1.548 2.583 19.88 0.174 0.437 −13.01
73 1.030 77.6 1.97 121.35 1.587 2.856 18.00 0.179 0.229 −15.37
74 1.044 79.9 3.51 93.05 1.629 3.098 15.92 0.181 −0.002 −15.93
75 1.058 83.3 4.63 65.70 1.675 3.311 13.95 0.179 −0.227 −14.68
76 1.072 87.5 5.39 44.39 1.724 3.497 12.25 0.174 −0.422 −12.28
77 1.087 92.1 5.90 29.92 1.775 3.661 10.74 0.167 −0.578 −9.59
78 1.101 97.1 6.24 20.34 1.829 3.804 9.31 0.158 −0.697 −7.02
79 1.115 102.4 6.48 13.82 1.884 3.927 7.94 0.147 −0.779 −4.37
80 1.130 107.8 6.64 9.60 1.941 4.031 6.65 0.135 −0.822 −1.42
81 1.144 113.3 6.75 7.77 1.999 4.118 5.44 0.123 −0.820 1.72
82 1.158 118.8 6.86 7.86 2.059 4.187 4.29 0.112 −0.772 4.87
83 1.173 124.5 6.98 7.28 2.119 4.240 3.03 0.101 −0.680 7.92
84 1.187 130.3 7.07 3.78 2.180 4.274 1.33 0.092 −0.546 10.66
85 1.201 136.1 7.09 −0.95 2.241 4.278 −1.08 0.086 −0.375 12.64
86 1.215 141.9 7.04 −3.31 2.302 4.243 −4.36 0.082 −0.184 13.44
87 1.230 147.6 6.99 −3.75 2.363 4.154 −8.57 0.080 0.009 12.89
88 1.244 153.3 6.93 −8.63 2.421 3.998 −13.62 0.082 0.184 10.90
89 1.258 159.0 6.74 −23.94 2.477 3.764 −19.21 0.086 0.321 7.37
90 1.273 164.4 6.25 −48.50 2.529 3.448 −24.71 0.091 0.395 2.47
91 1.287 169.2 5.36 −75.19 2.576 3.057 −29.09 0.097 0.391 −3.11
92 1.301 173.2 4.10 −96.87 2.616 2.616 −31.47 0.102 0.306 −8.34
93 1.316 175.9 2.59 −110.55 2.650 2.157 −31.41 0.106 0.153 −12.08
94 1.330 177.4 0.94 −117.61 2.678 1.718 −28.90 0.107 −0.039 −13.34
95 1.344 177.5 −0.78 −118.51 2.700 1.331 −24.40 0.105 −0.229 −11.78
96 1.358 176.1 −2.45 −109.12 2.716 1.020 −18.90 0.100 −0.376 −8.07
ISTUDY
HCR 97 1.373 173.5 −3.90 −84.77 2.729 0.790 −13.77 0.094 −0.460 −3.57
98 1.387 169.7 −4.88 −46.31 2.739 0.626 −10.14 0.087 −0.478 0.46
99 1.401 165.5 −5.22 −1.86 2.747 0.500 −8.25 0.080 −0.446 3.31
100 1.416 161.2 −4.93 36.59 2.753 0.390 −7.47 0.074 −0.384 4.77
101 1.430 157.4 −4.18 59.94 2.758 0.286 −7.03 0.069 −0.310 5.07
102 1.444 154.3 −3.22 66.68 2.761 0.189 −6.52 0.065 −0.239 4.60
103 1.459 152.1 −2.27 61.71 2.763 0.100 −5.67 0.062 −0.178 3.88
104 1.473 150.6 −1.45 51.41 2.764 0.027 −4.22 0.060 −0.128 3.28
105 1.487 149.7 −0.80 39.26 2.764 −0.021 −2.28 0.059 −0.084 2.79
106 1.501 149.3 −0.33 25.63 2.763 −0.038 −0.45 0.058 −0.048 2.13
ISTUDY
TABLE A.3b Linear and Angular Kinematics – Leg
TOR 1 0.000 39.8 −2.41 40.67 0.272 2.479 11.66 0.362 0.268 0.58
2 0.014 38.0 −1.70 56.27 0.308 2.618 7.99 0.366 0.277 0.37
3 0.029 37.0 −0.80 66.77 0.347 2.708 4.97 0.370 0.279 −0.29
4 0.043 36.7 0.21 71.22 0.386 2.760 2.66 0.374 0.268 −1.25
5 0.057 37.3 1.24 70.05 0.425 2.784 1.03 0.378 0.243 −2.27
6 0.072 38.8 2.21 64.35 0.465 2.789 −0.07 0.381 0.203 −3.27
7 0.086 41.0 3.08 55.75 0.505 2.782 −0.78 0.384 0.150 −4.09
8 0.100 43.8 3.81 46.45 0.545 2.767 −1.20 0.385 0.086 −4.55
9 0.114 47.2 4.41 38.02 0.584 2.748 −1.41 0.386 0.020 −4.61
10 0.129 51.0 4.90 30.50 0.624 2.727 −1.52 0.386 −0.045 −4.42
11 0.143 55.2 5.28 23.32 0.662 2.704 −1.61 0.385 −0.107 −4.07
12 0.157 59.7 5.56 16.47 0.701 2.681 −1.73 0.383 −0.162 −3.47
13 0.172 64.3 5.75 10.44 0.739 2.655 −1.99 0.380 −0.206 −2.63
14 0.186 69.1 5.86 5.57 0.777 2.624 −2.49 0.377 −0.237 −1.74
15 0.200 74.0 5.91 1.65 0.814 2.584 −3.23 0.374 −0.256 −0.98
16 0.215 78.8 5.91 −2.00 0.851 2.531 −4.22 0.370 −0.265 −0.37
17 0.229 83.6 5.85 −6.16 0.886 2.463 −5.38 0.366 −0.266 0.11
18 0.243 88.4 5.73 −11.89 0.921 2.378 −6.44 0.362 −0.262 0.45
19 0.257 93.0 5.51 −20.20 0.954 2.279 −7.20 0.358 −0.253 0.66
20 0.272 97.4 5.16 −31.54 0.986 2.171 −7.72 0.355 −0.243 0.87
21 0.286 101.5 4.61 −45.55 1.017 2.058 −8.18 0.352 −0.228 1.17
22 0.300 105.0 3.85 −60.60 1.045 1.938 −8.64 0.348 −0.209 1.35
23 0.315 107.8 2.88 −73.46 1.072 1.811 −8.98 0.346 −0.190 1.08
24 0.329 109.7 1.75 −80.46 1.097 1.681 −9.03 0.343 −0.178 0.43
25 0.343 110.7 0.58 −79.07 1.120 1.553 −8.67 0.340 −0.178 −0.16
26 0.357 110.7 −0.51 −68.79 1.141 1.433 −7.92 0.338 −0.183 −0.14
27 0.372 109.8 −1.39 −51.73 1.161 1.326 −6.93 0.335 −0.182 0.76
HCR 28 0.386 108.4 −1.99 −32.87 1.179 1.235 −5.87 0.333 −0.161 2.26
29 0.400 106.6 −2.33 −18.01 1.196 1.158 −4.90 0.331 −0.117 3.67
30 0.415 104.6 −2.50 −9.85 1.212 1.095 −4.10 0.329 −0.056 4.33
31 0.429 102.5 −2.61 −6.21 1.228 1.041 −3.70 0.329 −0.007 3.99
ISTUDY
32 0.443 100.3 −2.68 −2.78 1.242 0.989 −4.01 0.330 0.058 2.75
33 0.458 98.1 −2.69 2.90 1.256 0.926 −5.05 0.331 0.086 0.98
34 0.472 95.9 −2.60 10.00 1.269 0.844 −6.23 0.332 0.086 −0.67
35 0.486 93.8 −2.41 15.87 1.280 0.748 −6.81 0.333 0.066 −1.61
36 0.500 91.9 −2.15 18.47 1.290 0.650 −6.62 0.334 0.040 −1.74
37 0.515 90.3 −1.88 17.56 1.299 0.559 −6.09 0.334 0.016 −1.43
38 0.529 88.9 −1.64 14.51 1.306 0.475 −5.57 0.334 −0.001 −0.98
39 0.543 87.6 −1.46 10.97 1.312 0.399 −4.93 0.334 −0.012 −0.39
40 0.558 86.5 −1.33 7.87 1.317 0.334 −3.89 0.334 −0.012 0.40
41 0.572 85.4 −1.24 5.13 1.322 0.288 −2.49 0.334 0.000 1.14
42 0.586 84.4 −1.18 2.56 1.326 0.263 −1.08 0.334 0.020 1.31
43 0.601 83.5 −1.16 0.58 1.329 0.257 0.06 0.334 0.037 0.64
44 0.615 82.5 −1.17 −0.22 1.333 0.265 0.82 0.335 0.038 −0.53
45 0.629 81.6 −1.17 0.10 1.337 0.281 1.24 0.336 0.022 −1.50
46 0.643 80.6 −1.16 0.69 1.341 0.300 1.34 0.336 −0.005 −1.86
47 0.658 79.7 −1.15 0.65 1.345 0.319 1.27 0.335 −0.031 −1.62
48 0.672 78.7 −1.14 −0.22 1.350 0.336 1.21 0.335 −0.051 −0.95
49 0.686 77.8 −1.16 −1.19 1.355 0.353 1.17 0.334 −0.059 −0.09
50 0.701 76.8 −1.18 −1.64 1.360 0.370 1.00 0.333 −0.053 0.72
51 0.715 75.9 −1.20 −1.93 1.366 0.382 0.66 0.332 −0.038 1.30
52 0.729 74.9 −1.23 −2.99 1.371 0.389 0.45 0.332 −0.016 1.62
53 0.744 73.8 −1.29 −5.30 1.377 0.395 0.80 0.332 0.009 1.74
54 0.758 72.7 −1.39 −8.47 1.383 0.411 1.72 0.332 0.034 1.74
55 0.772 71.6 −1.53 −11.61 1.389 0.444 2.77 0.333 0.058 1.66
56 0.786 70.2 −1.72 −13.88 1.395 0.491 3.58 0.334 0.081 1.43
57 0.801 68.7 −1.93 −14.88 1.403 0.547 4.29 0.335 0.099 1.00
58 0.815 67.1 −2.14 −14.71 1.411 0.613 5.39 0.337 0.110 0.53
59 0.829 65.2 −2.35 −14.08 1.420 0.701 7.07 0.338 0.114 0.23
60 0.844 63.2 −2.55 −13.69 1.431 0.816 9.13 0.340 0.116 0.09
61 0.858 61.1 −2.74 −13.68 1.443 0.962 11.25 0.342 0.117 0.00
62 0.872 58.7 −2.94 −13.46 1.458 1.137 13.11 0.343 0.116 0.06
63 0.887 56.2 −3.13 −11.96 1.476 1.337 14.27 0.345 0.118 0.37
64 0.901 53.6 −3.28 −8.16 1.497 1.545 14.31 0.347 0.127 0.82
65 0.915 50.9 −3.36 −1.88 1.520 1.746 13.29 0.349 0.142 1.13
(continued )
ISTUDY
66 0.929 48.1 −3.33 6.33 1.547 1.926 11.68 0.351 0.159 1.26
67 0.944 45.4 −3.18 15.67 1.575 2.080 9.99 0.353 0.178 1.39
68 0.958 42.9 −2.88 25.25 1.606 2.211 8.47 0.356 0.199 1.66
69 0.972 40.7 −2.46 34.20 1.639 2.322 7.11 0.359 0.225 1.88
TOR 70 0.987 38.9 −1.91 42.20 1.672 2.415 5.93 0.362 0.253 1.75
71 1.001 37.6 −1.25 49.54 1.708 2.492 5.10 0.366 0.275 1.03
72 1.015 36.8 −0.49 56.03 1.744 2.560 4.60 0.370 0.282 −0.23
73 1.030 36.8 0.35 60.38 1.781 2.624 4.17 0.374 0.269 −1.71
74 1.044 37.4 1.24 61.40 1.819 2.680 3.53 0.378 0.234 −2.97
75 1.058 38.8 2.11 58.96 1.857 2.725 2.64 0.381 0.184 −3.79
76 1.072 40.9 2.92 53.72 1.897 2.755 1.59 0.383 0.125 −4.25
77 1.087 43.6 3.65 46.69 1.936 2.770 0.57 0.385 0.062 −4.57
78 1.101 46.8 4.26 38.96 1.976 2.771 −0.31 0.385 −0.006 −4.80
79 1.115 50.6 4.76 31.29 2.016 2.761 −0.96 0.384 −0.075 −4.73
80 1.130 54.6 5.15 24.07 2.055 2.744 −1.39 0.383 −0.141 −4.26
81 1.144 59.0 5.45 17.67 2.094 2.722 −1.68 0.380 −0.197 −3.51
82 1.158 63.6 5.66 12.52 2.133 2.696 −1.93 0.377 −0.241 −2.66
83 1.173 68.3 5.81 8.73 2.171 2.667 −2.22 0.373 −0.273 −1.78
84 1.187 73.1 5.91 6.21 2.209 2.632 −2.66 0.369 −0.292 −0.96
85 1.201 78.0 5.98 4.43 2.246 2.590 −3.35 0.365 −0.301 −0.35
86 1.215 82.9 6.04 1.90 2.283 2.536 −4.35 0.361 −0.302 0.05
87 1.230 87.9 6.04 −3.69 2.319 2.466 −5.61 0.356 −0.299 0.43
88 1.244 92.8 5.93 −14.21 2.354 2.376 −6.98 0.352 −0.290 0.92
89 1.258 97.6 5.63 −29.76 2.387 2.266 −8.35 0.348 −0.273 1.38
90 1.273 102.0 5.08 −48.61 2.418 2.137 −9.47 0.344 −0.250 1.67
91 1.287 105.9 4.24 −67.64 2.448 1.996 −9.96 0.341 −0.225 1.72
92 1.301 109.0 3.14 −82.66 2.476 1.853 −9.67 0.338 −0.201 1.54
93 1.316 111.0 1.88 −89.51 2.501 1.719 −8.82 0.335 −0.181 1.24
94 1.330 112.0 0.58 −85.71 2.525 1.600 −7.71 0.333 −0.166 1.20
95 1.344 112.0 −0.57 −71.95 2.547 1.499 −6.53 0.331 −0.147 1.66
96 1.358 111.1 −1.47 −52.52 2.568 1.414 −5.46 0.329 −0.118 2.38
ISTUDY
HCR 97 1.373 109.6 −2.08 −33.63 2.587 1.343 −4.69 0.327 −0.079 2.96
98 1.387 107.7 −2.43 −20.13 2.606 1.279 −4.45 0.326 −0.033 3.19
99 1.401 105.6 −2.65 −12.52 2.624 1.215 −4.84 0.326 0.012 2.88
100 1.416 103.4 −2.79 −7.44 2.641 1.141 −5.82 0.327 0.049 1.96
101 1.430 101.0 −2.86 −1.54 2.656 1.049 −7.22 0.328 0.068 0.66
102 1.444 98.7 −2.84 5.68 2.671 0.934 −8.71 0.329 0.068 −0.54
103 1.459 96.4 −2.70 12.64 2.683 0.800 −9.67 0.330 0.053 −1.23
104 1.473 94.2 −2.48 18.05 2.694 0.658 −9.59 0.330 0.033 −1.29
105 1.487 92.3 −2.19 21.86 2.702 0.526 −8.48 0.330 0.016 −0.96
106 1.501 90.7 −1.85 24.55 2.709 0.415 −6.99 0.331 0.005 −0.61
ISTUDY
TABLE A.3c Linear and Angular Kinematics – Thigh
TOR 1 0.000 82.7 3.29 24.42 0.430 2.081 1.36 0.652 −0.052 4.40
2 0.014 85.5 3.59 18.34 0.460 2.073 −1.89 0.652 0.026 6.06
3 0.029 88.6 3.81 12.48 0.489 2.027 −3.83 0.653 0.122 6.86
4 0.043 91.8 3.95 5.97 0.518 1.963 −4.53 0.655 0.222 6.72
5 0.057 95.1 3.98 −1.93 0.546 1.898 −4.32 0.659 0.314 5.74
6 0.072 98.3 3.89 −10.64 0.572 1.840 −3.64 0.664 0.386 4.11
7 0.086 101.4 3.68 −18.57 0.598 1.793 −2.92 0.670 0.432 2.20
8 0.100 104.3 3.36 −24.25 0.624 1.756 −2.45 0.677 0.449 0.42
9 0.114 106.9 2.99 −27.38 0.648 1.724 −2.28 0.683 0.444 −1.15
10 0.129 109.2 2.58 −28.82 0.673 1.691 −2.31 0.689 0.417 −2.60
11 0.143 111.2 2.16 −29.42 0.697 1.658 −2.41 0.695 0.369 −3.94
12 0.157 112.8 1.74 −29.22 0.720 1.622 −2.53 0.700 0.304 −4.97
13 0.172 114.0 1.33 −27.99 0.743 1.585 −2.69 0.704 0.227 −5.63
14 0.186 114.9 0.94 −26.18 0.766 1.545 −2.83 0.706 0.143 −6.02
15 0.200 115.6 0.58 −24.68 0.787 1.504 −2.79 0.708 0.055 −6.22
16 0.215 115.9 0.23 −23.85 0.809 1.466 −2.51 0.708 −0.035 −6.23
17 0.229 115.9 −0.11 −23.01 0.829 1.433 −2.04 0.707 −0.123 −5.97
18 0.243 115.7 −0.43 −20.96 0.850 1.407 −1.40 0.704 −0.206 −5.34
19 0.257 115.2 −0.70 −16.94 0.870 1.393 −0.55 0.701 −0.276 −4.22
20 0.272 114.6 −0.91 −11.04 0.889 1.392 0.42 0.696 −0.326 −2.51
21 0.286 113.7 −1.02 −4.04 0.909 1.405 1.38 0.692 −0.348 −0.34
22 0.300 112.9 −1.03 2.83 0.930 1.431 2.28 0.687 −0.336 1.68
23 0.315 112.1 −0.94 8.12 0.950 1.470 3.03 0.682 −0.300 2.98
24 0.329 111.3 −0.79 10.77 0.972 1.518 3.42 0.678 −0.251 3.50
25 0.343 110.8 −0.63 10.49 0.994 1.568 3.26 0.675 −0.200 3.62
26 0.357 110.3 −0.49 7.65 1.016 1.611 2.48 0.672 −0.148 3.69
27 0.372 110.0 −0.41 3.48 1.040 1.639 1.20 0.671 −0.094 3.92
HCR 28 0.386 109.6 −0.39 0.13 1.063 1.645 −0.31 0.670 −0.036 4.29
29 0.400 109.3 −0.41 −0.49 1.087 1.630 −1.67 0.669 0.029 4.62
30 0.415 109.0 −0.41 1.69 1.110 1.597 −2.61 0.670 0.096 4.64
31 0.429 108.6 −0.36 4.27 1.133 1.555 −3.28 0.672 0.161 4.04
ISTUDY
32 0.443 108.4 −0.29 3.58 1.154 1.503 −4.05 0.675 0.212 2.63
33 0.458 108.2 −0.26 −2.78 1.175 1.439 −4.94 0.678 0.236 0.62
34 0.472 107.9 −0.37 −13.38 1.196 1.362 −5.49 0.682 0.230 −1.26
35 0.486 107.6 −0.64 −23.43 1.214 1.282 −5.38 0.685 0.200 −2.26
36 0.500 106.9 −1.04 −28.36 1.232 1.208 −4.85 0.687 0.165 −2.32
37 0.515 105.9 −1.45 −26.92 1.249 1.143 −4.43 0.690 0.134 −2.00
38 0.529 104.5 −1.81 −20.92 1.265 1.082 −4.34 0.691 0.108 −1.77
39 0.543 102.9 −2.05 −12.96 1.280 1.019 −4.36 0.693 0.083 −1.52
40 0.558 101.2 −2.18 −4.84 1.294 0.957 −4.13 0.694 0.064 −1.01
41 0.572 99.3 −2.19 2.16 1.307 0.901 −3.50 0.695 0.054 −0.44
42 0.586 97.6 −2.11 6.84 1.320 0.857 −2.56 0.695 0.052 −0.31
43 0.601 95.9 −1.99 8.52 1.332 0.828 −1.53 0.696 0.045 −0.97
44 0.615 94.3 −1.87 7.81 1.344 0.813 −0.65 0.697 0.024 −2.12
45 0.629 92.8 −1.77 6.10 1.355 0.809 −0.07 0.697 −0.015 −2.99
46 0.643 91.4 −1.70 4.22 1.367 0.811 0.26 0.696 −0.062 −3.02
47 0.658 90.0 −1.65 2.22 1.378 0.817 0.61 0.695 −0.102 −2.35
48 0.672 88.7 −1.63 0.01 1.390 0.828 1.15 0.693 −0.129 −1.46
49 0.686 87.4 −1.65 −2.10 1.402 0.850 1.70 0.691 −0.143 −0.63
50 0.701 86.0 −1.69 −3.38 1.414 0.877 1.85 0.689 −0.147 0.01
51 0.715 84.6 −1.75 −3.24 1.427 0.902 1.51 0.687 −0.143 0.45
52 0.729 83.1 −1.79 −1.71 1.440 0.920 1.16 0.685 −0.134 0.67
53 0.744 81.7 −1.79 0.98 1.453 0.936 1.38 0.683 −0.124 0.79
54 0.758 80.2 −1.76 4.63 1.467 0.960 2.18 0.681 −0.112 0.78
55 0.772 78.8 −1.66 8.63 1.481 0.998 3.02 0.680 −0.101 0.58
56 0.786 77.5 −1.51 11.95 1.495 1.046 3.58 0.679 −0.095 0.20
57 0.801 76.3 −1.32 13.92 1.511 1.100 4.10 0.677 −0.096 −0.19
58 0.815 75.3 −1.11 14.82 1.527 1.163 5.02 0.676 −0.101 −0.46
59 0.829 74.5 −0.90 15.41 1.544 1.244 6.42 0.674 −0.109 −0.73
60 0.844 73.8 −0.67 16.49 1.562 1.347 7.98 0.673 −0.121 −1.12
61 0.858 73.4 −0.43 18.57 1.583 1.472 9.40 0.671 −0.141 −1.49
62 0.872 73.1 −0.14 21.54 1.605 1.616 10.36 0.669 −0.164 −1.58
63 0.887 73.1 0.19 24.59 1.629 1.768 10.40 0.666 −0.186 −1.28
64 0.901 73.4 0.56 27.13 1.655 1.913 9.09 0.663 −0.201 −0.68
65 0.915 74.1 0.97 29.35 1.683 2.028 6.50 0.660 −0.205 0.06
(continued )
ISTUDY
TABLE A.3c (Continued)
66 0.929 75.0 1.40 31.19 1.713 2.099 3.27 0.658 −0.199 0.94
67 0.944 76.4 1.86 31.71 1.743 2.122 0.24 0.655 −0.179 2.00
68 0.958 78.1 2.31 30.36 1.774 2.106 −2.10 0.652 −0.142 3.27
69 0.972 80.2 2.73 27.73 1.804 2.062 −3.60 0.651 0.085 4.59
TOR 70 0.987 82.5 3.10 24.46 1.833 2.003 −4.23 0.650 −0.010 5.65
71 1.001 85.2 3.43 20.30 1.861 1.941 −3.96 0.650 0.076 6.21
72 1.015 88.2 3.68 14.67 1.888 1.889 −3.05 0.652 0.167 6.16
73 1.030 91.3 3.85 7.52 1.915 1.854 −2.02 0.655 0.253 5.61
74 1.044 94.5 3.90 −0.87 1.941 1.832 −1.28 0.659 0.328 4.76
75 1.058 97.7 3.82 −10.14 1.967 1.817 −0.93 0.665 0.389 3.66
76 1.072 100.7 3.61 −19.00 1.993 1.805 −0.88 0.671 0.432 2.15
77 1.087 103.6 3.28 −25.45 2.019 1.792 −1.09 0.677 0.451 0.23
78 1.101 106.1 2.88 −28.34 2.045 1.774 −1.51 0.683 0.439 −1.77
79 1.115 108.3 2.47 −27.92 2.070 1.749 −2.07 0.689 0.400 −3.40
80 1.130 110.1 2.08 −25.47 2.095 1.715 −2.71 0.695 0.342 −4.45
81 1.144 111.7 1.74 −22.64 2.119 1.671 −3.30 0.699 0.273 −5.04
82 1.158 113.0 1.43 −20.83 2.142 1.620 −3.72 0.703 0.198 −5.39
83 1.173 114.0 1.14 −20.75 2.165 1.565 −3.87 0.705 0.119 −5.64
84 1.187 114.9 0.84 −22.11 2.187 1.510 −3.74 0.706 0.036 −5.90
85 1.201 115.4 0.51 −23.89 2.208 1.458 −3.37 0.706 −0.050 −6.16
86 1.215 115.7 0.16 −24.95 2.229 1.413 −2.72 0.705 −0.140 −6.20
87 1.230 115.7 −0.20 −24.36 2.249 1.380 −1.76 0.702 −0.227 −5.66
88 1.244 115.4 −0.54 −21.36 2.268 1.363 −0.58 0.698 −0.302 −4.43
89 1.258 114.8 −0.81 −15.42 2.288 1.364 0.58 0.693 −0.354 −2.69
90 1.273 114.0 −0.98 −6.72 2.307 1.380 1.61 0.688 −0.379 −0.70
91 1.287 113.2 −1.01 3.05 2.327 1.410 2.71 0.682 −0.374 1.30
92 1.301 112.4 −0.89 10.92 2.348 1.457 3.92 0.677 −0.342 3.07
93 1.316 111.7 −0.69 14.55 2.369 1.522 4.86 0.673 −0.287 4.38
94 1.330 111.3 −0.48 13.64 2.391 1.596 5.04 0.669 −0.216 5.19
95 1.344 110.9 −0.30 9.66 2.415 1.666 4.28 0.667 −0.138 5.50
96 1.358 110.8 −0.20 5.05 2.439 1.718 2.72 0.665 −0.059 5.33
ISTUDY
HCR 97 1.373 110.6 −0.16 2.00 2.464 1.744 0.71 0.665 0.014 4.88
98 1.387 110.5 −0.14 1.10 2.489 1.739 −1.30 0.666 0.080 4.42
99 1.401 110.4 −0.13 0.73 2.513 1.706 −3.03 0.667 0.141 3.89
100 1.416 110.3 −0.12 −1.58 2.537 1.652 −4.52 0.670 0.192 2.92
101 1.430 110.2 −0.17 −7.15 2.561 1.577 −5.90 0.673 0.224 1.39
102 1.444 110.0 −0.33 −14.74 2.583 1.483 −7.13 0.676 0.231 −0.34
103 1.459 109.6 −0.59 −21.06 2.603 1.373 −8.01 0.679 0.215 −1.65
104 1.473 109.0 −0.93 −22.96 2.622 1.254 −8.45 0.682 0.184 −2.18
105 1.487 108.1 −1.25 −19.15 2.639 1.132 −8.72 0.684 0.152 −2.11
106 1.501 107.0 −1.48 −10.16 2.654 1.005 −9.40 0.686 0.124 −1.89
ISTUDY
TABLE A.3d Linear and Angular Kinematics – ½ HAT
TOR 1 0.000 85.1 0.89 −11.61 0.473 1.377 1.23 1.080 0.035 2.76
2 0.014 85.7 0.71 −13.17 0.492 1.379 −0.70 1.081 0.084 3.89
3 0.029 86.2 0.52 −11.89 0.512 1.357 −2.11 1.083 0.146 4.49
4 0.043 86.6 0.37 −8.26 0.531 1.319 −2.87 1.085 0.212 4.46
5 0.057 86.8 0.28 −3.23 0.550 1.275 −2.98 1.089 0.274 3.80
6 0.072 87.0 0.27 1.64 0.568 1.234 −2.50 1.093 0.321 2.60
7 0.086 87.3 0.33 4.85 0.585 1.203 −1.64 1.098 0.348 1.19
8 0.100 87.6 0.41 6.05 0.602 1.187 −0.74 1.103 0.355 −0.18
9 0.114 88.0 0.50 6.17 0.619 1.182 −0.15 1.108 0.343 −1.50
10 0.129 88.4 0.59 6.19 0.636 1.183 0.06 1.113 0.312 −2.81
11 0.143 88.9 0.68 6.10 0.653 1.184 0.06 1.117 0.262 −3.89
12 0.157 89.5 0.76 5.21 0.670 1.184 0.06 1.120 0.201 −4.54
13 0.172 90.2 0.83 3.01 0.687 1.186 0.21 1.123 0.133 −4.81
14 0.186 90.9 0.85 −0.12 0.704 1.190 0.56 1.124 0.063 −4.87
15 0.200 91.6 0.82 −3.16 0.721 1.201 1.15 1.124 −0.007 −4.83
16 0.215 92.2 0.76 −5.47 0.738 1.223 1.93 1.124 −0.075 −4.72
17 0.229 92.8 0.67 −6.91 0.756 1.257 2.71 1.122 −0.142 −4.48
18 0.243 93.3 0.56 −7.30 0.774 1.300 3.24 1.120 −0.203 −3.98
19 0.257 93.7 0.46 −6.57 0.793 1.349 3.38 1.116 −0.255 −3.07
20 0.272 94.1 0.37 −4.89 0.813 1.397 3.11 1.112 −0.291 −1.63
21 0.286 94.3 0.32 −2.38 0.833 1.438 2.48 1.108 −0.302 0.10
22 0.300 94.6 0.31 0.85 0.854 1.468 1.66 1.104 −0.288 1.53
23 0.315 94.8 0.34 3.98 0.875 1.486 0.95 1.100 −0.259 2.21
24 0.329 95.1 0.42 5.39 0.896 1.495 0.64 1.096 −0.225 2.40
25 0.343 95.5 0.50 3.95 0.918 1.504 0.91 1.093 −0.190 2.60
26 0.357 96.0 0.53 0.20 0.939 1.521 1.52 1.091 −0.151 3.07
27 0.372 96.4 0.50 −4.14 0.961 1.548 1.98 1.089 −0.102 3.79
HCR 28 0.386 96.8 0.41 −7.80 0.984 1.578 1.91 1.088 −0.042 4.62
29 0.400 97.1 0.28 −10.40 1.006 1.602 1.32 1.088 0.030 5.28
30 0.415 97.2 0.12 −11.97 1.029 1.616 0.50 1.089 0.109 5.48
31 0.429 97.3 −0.06 −12.71 1.053 1.617 −0.32 1.091 0.186 4.93
ISTUDY
32 0.443 97.1 −0.25 −12.51 1.076 1.606 −1.04 1.094 0.250 3.50
33 0.458 96.9 −0.42 −10.48 1.098 1.587 −1.63 1.098 0.287 1.45
34 0.472 96.4 −0.55 −6.07 1.121 1.560 −1.99 1.102 0.291 −0.50
35 0.486 96.0 −0.59 −0.55 1.143 1.530 −2.10 1.106 0.272 −1.54
36 0.500 95.5 −0.56 3.64 1.165 1.500 −2.10 1.110 0.247 −1.63
37 0.515 95.0 −0.49 4.75 1.186 1.470 −2.14 1.114 0.226 −1.49
38 0.529 94.7 −0.43 2.67 1.207 1.439 −2.14 1.117 0.205 −1.69
39 0.543 94.3 −0.41 −1.41 1.227 1.409 −1.93 1.119 0.178 −1.99
40 0.558 94.0 −0.47 −5.74 1.247 1.384 −1.49 1.122 0.148 −1.98
41 0.572 93.6 −0.58 −8.82 1.267 1.366 −0.95 1.124 0.121 −1.69
42 0.586 93.0 −0.72 −9.90 1.286 1.356 −0.43 1.125 0.099 −1.57
43 0.601 92.4 −0.86 −9.06 1.305 1.354 0.04 1.126 0.076 −1.99
44 0.615 91.6 −0.98 −7.22 1.325 1.357 0.42 1.127 0.042 −2.88
45 0.629 90.8 −1.07 −5.50 1.344 1.366 0.71 1.128 −0.006 −3.64
46 0.643 89.9 −1.14 −4.17 1.364 1.378 0.90 1.127 −0.062 −3.69
47 0.658 88.9 −1.19 −2.51 1.384 1.392 1.02 1.126 −0.112 −3.04
48 0.672 87.9 −1.21 0.15 1.404 1.407 1.07 1.124 −0.149 −2.20
49 0.686 87.0 −1.18 3.16 1.424 1.422 1.00 1.122 −0.175 −1.53
50 0.701 86.0 −1.12 4.79 1.444 1.435 0.80 1.119 −0.192 −1.05
51 0.715 85.1 −1.05 4.15 1.465 1.445 0.57 1.116 −0.205 −0.71
52 0.729 84.3 −1.00 2.30 1.486 1.452 0.52 1.113 −0.213 −0.43
53 0.744 83.5 −0.98 1.15 1.506 1.460 0.71 1.110 −0.217 −0.07
54 0.758 82.7 −0.97 1.51 1.527 1.472 0.94 1.107 −0.215 0.31
55 0.772 81.9 −0.94 2.94 1.549 1.487 0.88 1.104 −0.208 0.52
56 0.786 81.2 −0.88 5.03 1.570 1.497 0.44 1.101 −0.200 0.55
57 0.801 80.5 −0.79 7.77 1.591 1.499 −0.02 1.098 −0.192 0.60
58 0.815 79.9 −0.66 10.87 1.613 1.497 −0.02 1.096 −0.183 0.79
59 0.829 79.4 −0.48 13.59 1.634 1.499 0.59 1.093 −0.169 0.91
60 0.844 79.1 −0.27 15.25 1.656 1.513 1.59 1.091 −0.156 0.79
61 0.858 78.9 −0.05 15.83 1.677 1.544 2.58 1.089 −0.147 0.58
62 0.872 79.0 0.18 15.92 1.700 1.587 3.10 1.086 −0.140 0.55
63 0.887 79.2 0.41 15.99 1.723 1.633 2.64 1.085 −0.131 0.71
(continued )
ISTUDY
TABLE A.3d (Continued)
64 0.901 79.7 0.64 15.90 1.747 1.663 0.88 1.083 −0.119 0.95
65 0.915 80.3 0.86 15.03 1.770 1.658 −1.90 1.081 −0.104 1.21
66 0.929 81.1 1.07 12.34 1.794 1.608 −4.71 1.080 −0.085 1.50
67 0.944 82.0 1.22 6.89 1.816 1.523 −6.30 1.079 −0.061 1.88
68 0.958 83.1 1.27 −0.81 1.838 1.428 −6.22 1.078 −0.031 2.43
69 0.972 84.1 1.19 −8.51 1.857 1.345 −5.08 1.078 0.008 3.16
TOR 70 0.987 85.0 1.02 −13.55 1.876 1.283 −3.73 1.078 0.059 3.86
71 1.001 85.8 0.81 −14.47 1.894 1.239 −2.53 1.079 0.119 4.26
72 1.015 86.3 0.61 −11.76 1.911 1.210 −1.56 1.082 0.181 4.25
73 1.030 86.8 0.47 −7.13 1.929 1.194 −0.86 1.085 0.240 3.91
74 1.044 87.1 0.40 −2.15 1.946 1.185 −0.45 1.089 0.293 3.44
75 1.058 87.4 0.41 2.49 1.962 1.181 −0.25 1.093 0.339 2.73
76 1.072 87.8 0.48 6.25 1.979 1.178 −0.13 1.098 0.371 1.47
77 1.087 88.2 0.59 8.26 1.996 1.177 0.01 1.104 0.381 −0.38
78 1.101 88.8 0.71 7.82 2.013 1.178 0.23 1.109 0.360 −2.42
79 1.115 89.4 0.81 4.89 2.030 1.184 0.59 1.114 0.312 −4.04
80 1.130 90.1 0.85 0.35 2.047 1.195 1.00 1.118 0.245 −4.95
81 1.144 90.8 0.82 −4.02 2.064 1.212 1.25 1.121 0.170 −5.20
82 1.158 91.4 0.74 −6.62 2.082 1.231 1.20 1.123 0.096 −5.05
83 1.173 92.0 0.63 −7.20 2.099 1.247 0.97 1.124 0.026 −4.71
84 1.187 92.5 0.53 −6.49 2.117 1.259 0.81 1.124 −0.039 −4.37
85 1.201 92.9 0.45 −5.14 2.135 1.270 0.83 1.123 −0.099 −4.19
86 1.215 93.2 0.38 −3.23 2.154 1.283 1.00 1.121 −0.159 −4.11
87 1.230 93.5 0.35 −0.73 2.172 1.299 1.26 1.118 −0.217 −3.86
88 1.244 93.8 0.36 1.86 2.191 1.319 1.47 1.115 −0.269 −3.18
89 1.258 94.1 0.41 3.58 2.210 1.341 1.52 1.110 −0.308 −2.09
90 1.273 94.4 0.47 3.89 2.229 1.362 1.48 1.106 −0.329 −0.83
91 1.287 94.9 0.52 3.56 2.249 1.383 1.53 1.101 −0.332 0.43
ISTUDY
92 1.301 95.3 0.57 3.94 2.269 1.406 1.67 1.096 −0.317 1.62
93 1.316 95.8 0.63 5.06 2.289 1.431 1.80 1.092 −0.285 2.69
94 1.330 96.3 0.71 5.42 2.309 1.457 1.92 1.088 −0.240 3.68
95 1.344 96.9 0.79 3.53 2.330 1.486 2.11 1.085 −0.180 4.53
96 1.358 97.6 0.81 −0.84 2.352 1.517 2.27 1.083 −0.110 5.12
HCR 97 1.373 98.3 0.76 −6.35 2.374 1.551 2.12 1.082 −0.034 5.43
98 1.387 98.9 0.63 −11.08 2.396 1.578 1.49 1.082 0.045 5.56
99 1.401 99.3 0.45 −13.91 2.419 1.593 0.57 1.083 0.125 5.41
100 1.416 99.6 0.23 −14.78 2.442 1.594 −0.38 1.086 0.200 4.65
101 1.430 99.7 0.02 −13.99 2.465 1.582 −1.19 1.089 0.258 3.12
102 1.444 99.6 −0.17 −11.75 2.487 1.560 −1.90 1.093 0.289 1.13
103 1.459 99.4 −0.31 −8.21 2.509 1.528 −2.78 1.097 0.291 −0.64
104 1.473 99.1 −0.40 −3.79 2.531 1.481 −4.11 1.101 0.271 −1.69
105 1.487 98.8 −0.42 0.37 2.552 1.410 −6.06 1.105 0.242 −2.07
106 1.501 98.4 −0.39 2.87 2.571 1.308 −8.68 1.108 0.212 −2.24
ISTUDY
TABLE A.4 Relative Joint Angular Kinematics – Ankle, Knee, and Hip
Ankle Knee Hip
Frame Time Theta Omega Alpha Theta Omega Alpha Theta Omega Alpha
s Deg R/s R/s/s Deg R/s R/s/s Deg R/s R/s/s
TOR 1 0.000 −15.2 −2.29 94.89 46.7 6.74 −21.91 −2.4 2.39 36.03
2 0.014 −16.4 −0.82 98.72 52.1 6.23 −46.59 −0.2 2.89 31.50
3 0.029 −16.5 0.54 84.63 56.9 5.41 −65.29 2.3 3.30 24.37
4 0.043 −15.6 1.60 63.13 61.0 4.37 −77.66 5.2 3.58 14.23
5 0.057 −13.9 2.34 41.31 64.1 3.19 −84.77 8.2 3.70 1.30
6 0.072 −11.7 2.78 20.10 66.2 1.94 −87.62 11.3 3.62 −12.28
7 0.086 −9.3 2.92 −0.74 67.3 0.68 −86.54 14.2 3.35 −23.42
8 0.100 −6.9 2.76 −18.70 67.3 −0.53 −81.99 16.8 2.95 −30.30
9 0.114 −4.8 2.38 −30.23 66.4 −1.66 −75.17 19.0 2.48 −33.55
10 0.129 −3.0 1.90 −34.13 64.6 −2.68 −67.64 20.8 1.99 −35.01
11 0.143 −1.7 1.41 −32.12 62.0 −3.60 −60.31 22.2 1.48 −35.53
12 0.157 −0.7 0.98 −26.84 58.7 −4.41 −53.09 23.3 0.98 −34.42
13 0.172 −0.1 0.64 −20.23 54.8 −5.11 −45.78 23.8 0.50 −30.99
14 0.186 0.3 0.40 −13.59 50.3 −5.72 −38.95 24.1 0.09 −26.06
15 0.200 0.6 0.25 −8.03 45.4 −6.23 −32.91 24.0 −0.25 −21.52
16 0.215 0.7 0.17 −3.87 40.1 −6.66 −26.57 23.7 −0.53 −18.38
17 0.229 0.9 0.14 −0.56 34.5 −6.99 −18.05 23.1 −0.77 −16.10
18 0.243 1.0 0.15 2.33 28.7 −7.17 −5.87 22.4 −0.99 −13.65
19 0.257 1.1 0.21 4.60 22.8 −7.16 10.65 21.5 −1.16 −10.37
20 0.272 1.3 0.29 5.47 16.9 −6.87 31.09 20.5 −1.28 −6.16
21 0.286 1.6 0.36 3.83 11.5 −6.27 53.59 19.4 −1.34 −1.66
22 0.300 1.9 0.40 −0.84 6.7 −5.34 74.76 18.3 −1.33 1.98
23 0.315 2.2 0.34 −8.36 2.7 −4.13 90.62 17.2 −1.28 4.14
24 0.329 2.5 0.16 −19.22 −0.1 −2.75 97.96 16.2 −1.21 5.38
25 0.343 2.5 −0.21 −33.28 −1.8 −1.33 95.28 15.2 −1.13 6.53
26 0.357 2.1 −0.79 −46.39 −2.3 −0.02 82.64 14.3 −1.03 7.45
27 0.372 1.2 −1.54 −50.65 −1.8 1.04 62.40 13.6 −0.92 7.62
HCR 28 0.386 −0.4 −2.24 −38.30 −0.6 1.76 40.55 12.8 −0.81 7.93
29 0.400 −2.5 −2.63 −7.99 1.1 2.20 24.37 12.2 −0.69 9.91
30 0.415 −4.7 −2.47 30.12 3.0 2.46 16.62 11.7 −0.52 13.66
ISTUDY
31 0.429 −6.5 −1.77 59.60 5.1 2.67 12.93 11.4 −0.30 16.99
32 0.443 −7.6 −0.77 70.70 7.4 2.83 5.97 11.2 −0.04 16.09
33 0.458 −7.8 0.25 64.54 9.8 2.84 −8.17 11.3 0.16 7.69
34 0.472 −7.2 1.08 46.87 12.1 2.60 −27.06 11.5 0.18 −7.32
35 0.486 −6.0 1.59 24.33 14.0 2.07 −44.04 11.6 −0.05 −22.88
36 0.500 −4.6 1.77 4.14 15.4 1.34 −53.13 11.4 −0.47 −32.00
37 0.515 −3.1 1.71 −8.73 16.2 0.55 −52.57 10.8 −0.96 −31.67
38 0.529 −1.8 1.52 −13.85 16.3 −0.17 −44.30 9.8 −1.38 −23.59
39 0.543 −0.6 1.31 −13.32 15.9 −0.72 −31.51 8.6 −1.64 −11.54
40 0.558 0.4 1.14 −9.71 15.2 −1.07 −17.12 7.2 −1.71 0.90
41 0.572 1.2 1.04 −6.42 14.2 −1.21 −3.98 5.8 −1.61 10.98
42 0.586 2.1 0.96 −6.79 13.2 −1.18 5.07 4.5 −1.40 16.74
43 0.601 2.8 0.84 −10.34 12.3 −1.06 8.50 3.5 −1.13 17.58
44 0.615 3.4 0.66 −12.36 11.4 −0.94 7.49 2.7 −0.89 15.03
45 0.629 3.9 0.49 −9.52 10.7 −0.85 4.99 2.0 −0.70 11.60
46 0.643 4.2 0.39 −3.26 10.1 −0.80 3.03 1.5 −0.56 8.39
47 0.658 4.6 0.40 3.33 9.4 −0.76 1.90 1.1 −0.46 4.74
48 0.672 4.9 0.49 7.14 8.8 −0.74 1.14 0.8 −0.43 −0.13
49 0.686 5.3 0.60 4.92 8.2 −0.73 0.45 0.4 −0.47 −5.26
50 0.701 5.9 0.63 −3.15 7.6 −0.73 0.18 0.0 −0.58 −8.17
51 0.715 6.4 0.51 −11.50 7.0 −0.72 1.21 −0.5 −0.70 −7.39
52 0.729 6.7 0.30 −14.34 6.4 −0.69 4.06 −1.2 −0.79 −4.00
53 0.744 6.9 0.10 −10.68 5.9 −0.61 8.77 −1.8 −0.81 −0.17
54 0.758 6.9 −0.01 −4.17 5.4 −0.44 15.06 −2.5 −0.79 3.12
55 0.772 6.9 −0.02 0.00 5.2 −0.18 22.07 −3.1 −0.73 5.70
56 0.786 6.8 −0.01 −1.14 5.1 0.19 28.41 −3.7 −0.63 6.92
57 0.801 6.8 −0.05 −7.42 5.5 0.63 32.98 −4.2 −0.53 6.16
58 0.815 6.8 −0.22 −17.46 6.2 1.13 35.54 −4.6 −0.45 3.95
59 0.829 6.5 −0.55 −28.75 7.3 1.65 36.64 −4.9 −0.41 1.81
60 0.844 5.9 −1.04 −37.44 8.9 2.18 37.19 −5.2 −0.40 1.24
61 0.858 4.8 −1.62 −40.93 10.9 2.71 37.90 −5.6 −0.38 2.74
62 0.872 3.2 −2.21 −40.94 13.3 3.26 39.00 −5.9 −0.32 5.62
63 0.887 1.2 −2.79 −41.67 16.2 3.83 40.04 −6.1 −0.22 8.61
(continued )
ISTUDY
Ankle Knee Hip
Frame Time Theta Omega Alpha Theta Omega Alpha Theta Omega Alpha
s Deg R/s R/s/s Deg R/s R/s/s Deg R/s R/s/s
64 0.901 −1.4 −3.40 −43.21 19.6 4.41 39.95 −6.2 −0.08 11.23
65 0.915 −4.4 −4.03 −38.18 23.5 4.97 37.59 −6.2 0.10 14.31
66 0.929 −8.0 −4.50 −17.72 27.8 5.48 32.25 −6.0 0.33 18.85
67 0.944 −11.8 −4.54 18.41 32.4 5.90 23.44 −5.7 0.64 24.81
68 0.958 −15.4 −3.97 59.42 37.4 6.15 10.87 −5.0 1.04 31.17
69 0.972 −18.3 −2.84 91.15 42.5 6.21 −4.80 −4.0 1.53 36.24
TOR 70 0.987 −20.1 −1.36 104.21 47.6 6.02 −21.92 −2.5 2.08 38.01
71 1.001 −20.5 0.14 96.45 52.4 5.58 −38.75 −0.5 2.62 34.78
72 1.015 −19.8 1.40 72.74 56.7 4.91 −53.72 1.8 3.07 26.42
73 1.030 −18.2 2.22 42.50 60.4 4.04 −65.55 4.5 3.38 14.66
74 1.044 −16.2 2.61 15.01 63.4 3.03 −74.32 7.3 3.49 1.29
75 1.058 −14.0 2.65 −3.93 65.4 1.92 −80.71 10.2 3.41 −12.63
76 1.072 −11.8 2.50 −13.24 66.5 0.73 −83.96 12.9 3.13 −25.25
77 1.087 −9.9 2.28 −16.29 66.6 −0.48 −82.70 15.3 2.69 −33.71
78 1.101 −8.1 2.03 −17.51 65.7 −1.64 −76.99 17.3 2.17 −36.16
79 1.115 −6.5 1.77 −18.73 63.9 −2.69 −68.11 18.9 1.66 −32.81
80 1.130 −5.2 1.50 −19.65 61.3 −3.59 −57.74 20.1 1.23 −25.82
81 1.144 −4.1 1.21 −19.37 58.0 −4.34 −48.02 20.9 0.92 −18.62
82 1.158 −3.2 0.94 −17.21 54.2 −4.96 −40.91 21.6 0.70 −14.21
83 1.173 −2.5 0.72 −13.87 49.9 −5.51 −37.09 22.1 0.51 −13.55
84 1.187 −2.0 0.55 −10.79 45.2 −6.02 −35.35 22.4 0.31 −15.62
85 1.201 −1.6 0.41 −7.88 40.0 −6.52 −32.90 22.6 0.06 −18.75
86 1.215 −1.3 0.32 −3.38 34.5 −6.96 −26.57 22.5 −0.23 −21.73
87 1.230 −1.1 0.31 3.27 28.6 −7.28 −14.15 22.2 −0.56 −23.63
88 1.244 −0.8 0.42 9.14 22.6 −7.37 4.76 21.6 −0.90 −23.22
89 1.258 −0.4 0.58 10.06 16.5 −7.14 28.98 20.7 −1.22 −19.01
90 1.273 0.1 0.70 4.80 10.9 −6.54 56.01 19.6 −1.45 −10.61
ISTUDY
91 1.287 0.7 0.71 −4.67 5.8 −5.54 81.46 18.3 −1.52 −0.51
92 1.301 1.3 0.57 −16.27 1.8 −4.21 99.89 17.1 −1.46 6.98
93 1.316 1.7 0.25 −29.67 −1.1 −2.68 107.07 15.9 −1.32 9.49
94 1.330 1.7 −0.28 −44.20 −2.6 −1.15 101.58 14.9 −1.19 8.22
95 1.344 1.2 −1.02 −54.99 −3.0 0.22 85.39 14.0 −1.09 6.14
96 1.358 0.0 −1.85 −53.10 −2.3 1.30 63.88 13.1 −1.01 5.89
HCR 97 1.373 −1.8 −2.53 −31.93 −0.8 2.05 43.96 12.3 −0.92 8.35
98 1.387 −4.1 −2.77 4.96 1.1 2.55 29.90 11.6 −0.78 12.18
99 1.401 −6.4 −2.39 43.33 3.4 2.90 20.18 11.1 −0.57 14.65
100 1.416 −8.0 −1.53 67.62 5.9 3.13 9.21 10.7 −0.36 13.20
101 1.430 −8.9 −0.46 71.62 8.5 3.17 −6.47 10.5 −0.20 6.83
102 1.444 −8.8 0.52 59.88 11.0 2.94 −24.49 10.4 −0.16 −2.99
103 1.459 −8.0 1.25 41.13 13.3 2.47 −39.25 10.2 −0.28 −12.86
104 1.473 −6.7 1.70 22.10 15.1 1.82 −46.90 9.9 −0.53 −19.17
105 1.487 −5.2 1.89 5.23 16.3 1.13 −47.03 9.4 −0.83 −19.53
106 1.501 −3.7 1.85 −10.10 16.9 0.48 −40.67 8.5 −1.09 −13.02
ISTUDY
TABLE A.5a Reaction Forces and Moments of Force – Ankle and Knee
Foot Segment Leg Segment
Ground Ankle Ground Ankle Ankle Knee Ankle Knee
Frame Time RX RY RX RY Cof P X Moment RX RY RX RY Moment Moment
s N N N N m N m N N N N N m N m
TOR 1 0.000 0.0 0.0 20.9 3.9 0.000 1.6 −20.9 −3.9 52.0 31.6 −1.6 7.0
2 0.014 0.0 0.0 19.8 0.0 0.000 1.6 −19.8 0.0 41.1 27.1 −1.6 7.0
3 0.029 0.0 0.0 17.6 −3.1 0.000 1.5 −17.6 3.1 30.8 22.2 −1.5 6.9
4 0.043 0.0 0.0 14.9 −5.0 0.000 1.3 −14.9 5.0 22.0 17.8 −1.3 6.5
5 0.057 0.0 0.0 12.3 −5.5 0.000 1.1 −12.3 5.5 15.1 14.6 −1.1 5.8
6 0.072 0.0 0.0 10.3 −4.7 0.000 0.9 −10.3 4.7 10.1 12.7 −0.9 4.8
7 0.086 0.0 0.0 8.6 −3.0 0.000 0.7 −8.6 3.0 6.5 12.2 −0.7 3.6
8 0.100 0.0 0.0 7.3 −0.7 0.000 0.6 −7.3 0.7 4.1 13.4 −0.6 2.3
9 0.114 0.0 0.0 6.3 2.0 0.000 0.5 −6.3 −2.0 2.5 15.9 −0.5 1.0
10 0.129 0.0 0.0 5.5 4.8 0.000 0.4 −5.5 −4.8 1.4 19.1 −0.4 −0.1
11 0.143 0.0 0.0 4.7 7.5 0.000 0.4 −4.7 −7.5 0.4 22.8 −0.4 −1.0
12 0.157 0.0 0.0 3.7 10.2 0.000 0.5 −3.7 −10.2 −0.9 27.1 −0.5 −1.9
13 0.172 0.0 0.0 2.5 12.6 0.000 0.5 −2.5 −12.6 −2.8 31.8 −0.5 −2.7
14 0.186 0.0 0.0 0.9 14.7 0.000 0.6 −0.9 −14.7 −5.7 36.2 −0.6 −3.5
15 0.200 0.0 0.0 −0.9 16.1 0.000 0.6 0.9 −16.1 −9.5 39.6 −0.6 −4.2
16 0.215 0.0 0.0 −3.0 16.8 0.000 0.6 3.0 −16.8 −14.2 42.0 −0.6 −4.9
17 0.229 0.0 0.0 −5.4 17.0 0.000 0.6 5.4 −17.0 −19.7 43.4 −0.6 −5.8
18 0.243 0.0 0.0 −8.3 16.3 0.000 0.6 8.3 −16.3 −25.5 43.7 −0.6 −6.8
19 0.257 0.0 0.0 −11.5 14.8 0.000 0.5 11.5 −14.8 −30.7 42.7 −0.5 −8.0
20 0.272 0.0 0.0 −14.9 12.2 0.000 0.3 14.9 −12.2 −35.5 40.7 −0.3 −9.4
21 0.286 0.0 0.0 −18.1 9.0 0.000 0.1 18.1 −9.0 −39.9 38.2 −0.1 −11.1
22 0.300 0.0 0.0 −20.8 5.2 0.000 −0.1 20.8 −5.2 −43.9 34.9 0.1 −12.8
23 0.315 0.0 0.0 −22.7 1.6 0.000 −0.3 22.7 −1.6 −46.6 30.6 0.3 −14.3
24 0.329 0.0 0.0 −23.2 −1.2 0.000 −0.5 23.2 1.2 −47.3 26.1 0.5 −15.1
25 0.343 0.0 0.0 −22.2 −2.4 0.000 −0.5 22.2 2.4 −45.3 23.3 0.5 −14.6
26 0.357 0.0 0.0 −19.5 −1.7 0.000 −0.5 19.5 1.7 −40.6 24.1 0.5 −12.7
27 0.372 0.0 0.0 −15.6 1.0 0.000 −0.3 15.6 −1.0 −34.1 29.1 0.3 −9.5
ISTUDY
HCR 28 0.386 37.3 87.1 −48.7 −82.4 1.227 −1.7 48.7 82.4 −64.3 −50.2 1.7 −33.8
29 0.400 −4.2 192.6 −3.8 −184.1 1.244 4.6 3.8 184.1 −16.8 −148.2 −4.6 −19.9
30 0.415 −43.7 304.1 38.2 −292.9 1.261 8.3 −38.2 292.9 27.2 −255.2 −8.3 −7.6
31 0.429 −74.0 404.2 69.8 −391.4 1.291 2.8 −69.8 391.4 59.9 −354.6 −2.8 −4.5
32 0.443 −91.9 476.6 88.4 −463.4 1.290 6.9 −88.4 463.4 77.7 −429.9 −6.9 7.8
33 0.458 −102.7 521.7 99.7 −508.9 1.301 4.6 −99.7 508.9 86.3 −480.2 −4.6 14.5
34 0.472 −110.5 552.9 108.1 −541.1 1.304 5.0 −108.1 541.1 91.5 −516.7 −5.0 24.8
35 0.486 −114.2 579.8 112.5 −569.0 1.311 2.6 −112.5 569.0 94.3 −547.2 −2.6 31.6
36 0.500 −110.5 599.6 109.4 −589.9 1.317 −0.5 −109.4 589.9 91.7 −568.4 0.5 35.0
37 0.515 −98.1 604.5 97.1 −595.7 1.316 0.2 −97.1 595.7 80.9 −573.4 −0.2 37.8
38 0.529 −79.8 589.9 78.5 −581.7 1.321 −3.5 −78.5 581.7 63.7 −558.2 3.5 32.5
39 0.543 −62.0 558.1 60.3 −550.2 1.322 −5.1 −60.3 550.2 47.2 −525.1 5.1 28.1
40 0.558 −48.7 516.7 47.2 −508.7 1.325 −6.6 −47.2 508.7 36.8 −481.5 6.6 24.8
41 0.572 −39.8 473.4 38.8 −465.3 1.333 −10.5 −38.8 465.3 32.2 −436.1 10.5 20.2
42 0.586 −33.0 433.4 32.7 −425.2 1.335 −10.6 −32.7 425.2 29.8 −395.6 10.6 19.8
43 0.601 −27.0 400.3 27.2 −392.2 1.345 −14.1 −27.2 392.2 27.4 −364.4 14.1 15.8
44 0.615 −21.8 377.1 22.4 −369.4 1.358 −18.7 −22.4 369.4 24.6 −344.7 18.7 10.9
45 0.629 −18.1 365.0 19.0 −357.6 1.369 −22.5 −19.0 357.6 22.3 −335.4 22.5 7.8
46 0.643 −16.6 362.3 17.4 −355.0 1.377 −25.3 −17.4 355.0 21.0 −333.8 25.3 6.6
47 0.658 −16.8 366.5 17.6 −359.1 1.384 −27.9 −17.6 359.1 21.0 −337.2 27.9 6.6
48 0.672 −16.8 375.0 17.5 −367.3 1.390 −30.3 −17.5 367.3 20.7 −343.7 30.3 7.0
49 0.686 −14.5 386.1 15.2 −378.2 1.396 −33.5 −15.2 378.2 18.3 −352.3 33.5 5.9
50 0.701 −9.6 400.3 10.3 −392.4 1.406 −38.8 −10.3 392.4 13.0 −364.3 38.8 2.3
51 0.715 −3.8 418.8 4.2 −410.8 1.417 −45.1 −4.2 410.8 6.0 −381.2 45.1 −2.3
52 0.729 2.2 441.0 −2.1 −432.7 1.412 −45.4 2.1 432.7 −0.9 −402.2 45.4 −0.1
53 0.744 8.7 465.8 −9.0 −457.0 1.424 −53.5 9.0 457.0 −6.8 −426.3 53.5 −5.0
54 0.758 16.5 493.7 −16.8 −484.4 1.428 −58.8 16.8 484.4 −12.3 −453.7 58.8 −6.2
55 0.772 26.8 523.9 −27.1 −514.2 1.432 −64.8 27.1 514.2 −19.7 −483.6 64.8 −7.9
56 0.786 40.1 552.6 −40.4 −542.9 1.436 −71.2 40.4 542.9 −30.9 −513.0 71.2 −10.4
57 0.801 54.8 576.8 −55.0 −567.5 1.440 −77.3 55.0 567.5 −43.5 −538.7 77.3 −12.5
58 0.815 68.1 595.8 −67.6 −586.9 1.444 −82.7 67.6 586.9 −53.3 −559.3 82.7 −12.4
59 0.829 79.6 608.6 −77.9 −599.7 1.447 −87.0 77.9 599.7 −59.1 −573.0 87.0 −10.0
60 0.844 90.8 612.1 −87.4 −602.8 1.451 −89.7 87.4 602.8 −63.1 −576.4 89.7 −6.4
(continued )
ISTUDY
Foot Segment Leg Segment
Ground Ankle Ground Ankle Ankle Knee Ankle Knee
Frame Time RX RY RX RY Cof P X Moment RX RY RX RY Moment Moment
s N N N N m N m N N N N N m N m
61 0.858 101.5 602.3 −96.4 −592.2 1.455 −89.8 96.4 592.2 −66.5 −566.1 89.8 −2.2
62 0.872 110.4 576.1 −103.8 −565.0 1.459 −86.7 103.8 565.0 −68.9 −538.7 86.7 2.3
63 0.887 115.6 530.3 −107.7 −518.2 1.463 −79.7 107.7 518.2 −69.7 −491.1 79.7 6.5
64 0.901 114.5 463.0 −105.7 −450.2 1.467 −68.6 105.7 450.2 −67.5 −421.9 68.6 10.1
65 0.915 105.2 377.3 −95.5 −364.4 1.470 −54.3 95.5 364.4 −60.1 −335.2 54.3 12.5
66 0.929 88.2 282.1 −77.2 −269.9 1.474 −38.8 77.2 269.9 −46.1 −240.4 38.8 13.3
67 0.944 65.7 190.1 −53.0 −179.1 1.478 −24.2 53.0 179.1 −26.4 −149.3 24.2 12.5
68 0.958 41.4 110.8 −26.7 −101.8 1.482 −12.3 26.7 101.8 −4.1 −71.2 12.3 10.6
69 0.972 18.4 44.4 −2.1 −37.9 1.486 −3.6 2.1 37.9 16.9 −6.8 3.6 6.8
TOR 70 0.987 0.0 0.0 17.1 3.5 0.000 1.4 −17.1 −3.5 32.9 34.3 −1.4 3.6
71 1.001 0.0 0.0 16.8 0.3 0.000 1.4 −16.8 −0.3 30.4 29.2 −1.4 4.6
72 1.015 0.0 0.0 15.8 −2.5 0.000 1.4 −15.8 2.5 28.0 23.0 −1.4 5.7
73 1.030 0.0 0.0 14.3 −4.4 0.000 1.3 −14.3 4.4 25.4 17.2 −1.3 6.4
74 1.044 0.0 0.0 12.6 −4.9 0.000 1.1 −12.6 4.9 22.1 13.4 −1.1 6.3
75 1.058 0.0 0.0 11.1 −3.9 0.000 0.9 −11.1 3.9 18.1 12.2 −0.9 5.5
76 1.072 0.0 0.0 9.7 −2.0 0.000 0.7 −9.7 2.0 14.0 12.8 −0.7 4.3
77 1.087 0.0 0.0 8.5 0.2 0.000 0.6 −8.5 −0.2 10.0 14.1 −0.6 3.0
78 1.101 0.0 0.0 7.4 2.2 0.000 0.5 −7.4 −2.2 6.6 15.6 −0.5 1.8
79 1.115 0.0 0.0 6.3 4.3 0.000 0.5 −6.3 −4.3 3.7 17.8 −0.5 0.6
80 1.130 0.0 0.0 5.3 6.7 0.000 0.5 −5.3 −6.7 1.6 21.5 −0.5 −0.5
81 1.144 0.0 0.0 4.3 9.2 0.000 0.5 −4.3 −9.2 −0.2 25.9 −0.5 −1.5
82 1.158 0.0 0.0 3.4 11.7 0.000 0.6 −3.4 −11.7 −1.7 30.7 −0.6 −2.2
83 1.173 0.0 0.0 2.4 14.1 0.000 0.6 −2.4 −14.1 −3.5 35.5 −0.6 −2.7
84 1.187 0.0 0.0 1.1 16.3 0.000 0.7 −1.1 −16.3 −6.0 39.8 −0.7 −3.1
85 1.201 0.0 0.0 −0.9 17.8 0.000 0.8 0.9 −17.8 −9.8 43.0 −0.8 −3.6
86 1.215 0.0 0.0 −3.5 18.5 0.000 0.8 3.5 −18.5 −15.1 44.7 −0.8 −4.3
ISTUDY
87 1.230 0.0 0.0 −6.8 18.0 0.000 0.7 6.8 −18.0 −21.7 45.3 −0.7 −5.6
88 1.244 0.0 0.0 −10.8 16.4 0.000 0.6 10.8 −16.4 −29.4 45.0 −0.6 −7.4
89 1.258 0.0 0.0 −15.3 13.6 0.000 0.4 15.3 −13.6 −37.5 43.5 −0.4 −9.7
90 1.273 0.0 0.0 −19.6 9.7 0.000 0.1 19.6 −9.7 −44.8 40.3 −0.1 −12.3
91 1.287 0.0 0.0 −23.1 5.3 0.000 −0.2 23.1 −5.3 −49.6 36.0 0.2 −14.5
92 1.301 0.0 0.0 −25.0 1.2 0.000 −0.4 25.0 −1.2 −50.8 31.4 0.4 −15.9
93 1.316 0.0 0.0 −24.9 −1.8 0.000 −0.5 24.9 1.8 −48.4 27.6 0.5 −15.9
94 1.330 0.0 0.0 −22.9 −2.8 0.000 −0.6 22.9 2.8 −43.5 26.5 0.6 −14.5
95 1.344 0.0 0.0 −19.4 −1.6 0.000 −0.5 19.4 1.6 −36.8 29.0 0.5 −11.7
96 1.358 0.0 0.0 −15.0 1.4 0.000 −0.3 15.0 −1.4 −29.5 33.9 0.3 −8.2
HCR 97 1.373 0.0 0.0 −10.9 4.9 0.000 0.0 10.9 −4.9 −23.4 39.0 0.0 −4.9
98 1.387 0.0 0.0 −8.1 8.2 0.000 0.3 8.1 −8.2 −19.9 42.8 −0.3 −2.8
99 1.401 0.0 0.0 −6.5 10.4 0.000 0.5 6.5 −10.4 −19.4 44.2 −0.5 −2.0
100 1.416 0.0 0.0 −5.9 11.6 0.000 0.6 5.9 −11.6 −21.4 42.9 −0.6 −2.4
101 1.430 0.0 0.0 −5.6 11.8 0.000 0.7 5.6 −11.8 −24.8 39.7 −0.7 −3.2
102 1.444 0.0 0.0 −5.2 11.4 0.000 0.7 5.2 −11.4 −28.4 36.1 −0.7 −4.1
103 1.459 0.0 0.0 −4.5 10.9 0.000 0.6 4.5 −10.9 −30.3 33.7 −0.6 −4.6
104 1.473 0.0 0.0 −3.4 10.4 0.000 0.6 3.4 −10.4 −28.9 33.1 −0.6 −4.2
105 1.487 0.0 0.0 −1.8 10.0 0.000 0.6 1.8 −10.0 −24.4 33.6 −0.6 −3.2
106 1.501 0.0 0.0 −0.4 9.5 0.000 0.6 0.4 −9.5 −19.0 34.0 −0.6 −2.1
ISTUDY
TABLE A.5b Reaction Forces and Moments of Force – Hip
Thigh Segment
Knee Hip Knee Hip
Frame Time RX RY RX RY Moment Moment
s N N N N N m N m
TOR 1 0.000 −52.0 −31.6 59.7 112.1 −7.0 22.9
2 0.014 −41.1 −27.1 30.3 117.1 −7.0 17.8
3 0.029 −30.8 −22.2 9.1 116.7 −6.9 13.8
4 0.043 −22.0 −17.8 −3.8 111.5 −6.5 10.8
5 0.057 −15.1 −14.6 −9.4 102.8 −5.8 8.5
6 0.072 −10.1 −12.7 −10.6 91.7 −4.8 6.7
7 0.086 −6.5 −12.2 −10.0 80.4 −3.6 5.0
8 0.100 −4.1 −13.4 −9.8 71.3 −2.3 3.4
9 0.114 −2.5 −15.9 −10.4 65.0 −1.0 2.0
10 0.129 −1.4 −19.1 −11.7 60.0 0.1 0.9
11 0.143 −0.4 −22.8 −13.3 56.1 1.0 0.0
12 0.157 0.9 −27.1 −15.2 54.5 1.9 −0.8
13 0.172 2.8 −31.8 −18.1 55.5 2.7 −1.5
14 0.186 5.7 −36.2 −21.7 57.6 3.5 −2.5
15 0.200 9.5 −39.6 −25.3 59.9 4.2 −3.6
16 0.215 14.2 −42.0 −28.4 62.3 4.9 −5.0
17 0.229 19.7 −43.4 −31.3 65.1 5.8 −6.7
18 0.243 25.5 −43.7 −33.4 69.0 6.8 −8.7
19 0.257 30.7 −42.7 −33.9 74.4 8.0 −10.5
20 0.272 35.5 −40.7 −33.1 82.1 9.4 −12.2
21 0.286 39.9 −38.2 −32.1 91.9 11.1 −14.0
22 0.300 43.9 −34.9 −30.9 100.1 12.8 −16.0
23 0.315 46.6 −30.6 −29.4 103.1 14.3 −17.9
24 0.329 47.3 −26.1 −27.9 101.5 15.1 −19.1
25 0.343 45.3 −23.3 −26.8 99.4 14.6 −18.7
26 0.357 40.6 −24.1 −26.5 100.7 12.7 −16.1
27 0.372 34.1 −29.1 −27.3 106.9 9.5 −11.7
ISTUDY
HCR 28 0.386 64.3 50.2 −66.1 29.7 33.8 −54.4
29 0.400 16.8 148.2 −26.3 −66.4 19.9 −37.6
30 0.415 −27.2 255.2 12.4 −173.3 7.6 −23.5
31 0.429 −59.9 354.6 41.3 −276.1 4.5 −20.8
32 0.443 −77.7 429.9 54.7 −359.4 −7.8 −11.1
33 0.458 −86.3 480.2 58.3 −421.0 −14.5 −7.8
34 0.472 −91.5 516.7 60.4 −468.3 −24.8 −0.4
35 0.486 −94.3 547.2 63.8 −504.4 −31.6 4.7
36 0.500 −91.7 568.4 64.2 −526.0 −35.0 7.3
37 0.515 −80.9 573.4 55.7 −529.1 −37.8 9.9
38 0.529 −63.7 558.2 39.0 −512.7 −32.5 5.0
39 0.543 −47.2 525.1 22.5 −478.1 −28.1 3.0
40 0.558 −36.8 481.5 13.4 −431.6 −24.8 4.7
41 0.572 −32.2 436.1 12.3 −383.0 −20.2 6.5
42 0.586 −29.8 395.6 15.2 −341.7 −19.8 12.0
43 0.601 −27.4 364.4 18.7 −314.3 −15.8 12.5
44 0.615 −24.6 344.7 21.0 −301.1 −10.9 10.9
45 0.629 −22.3 335.4 21.9 −296.7 −7.8 10.1
46 0.643 −21.0 333.8 22.5 −295.3 −6.6 11.2
47 0.658 −21.0 337.2 24.4 −295.0 −6.6 13.8
48 0.672 −20.7 343.7 27.3 −296.3 −7.0 16.7
49 0.686 −18.3 352.3 27.9 −300.2 −5.9 17.7
50 0.701 −13.0 364.3 23.4 −308.6 −2.3 15.2
51 0.715 −6.0 381.2 14.5 −323.0 2.3 11.2
52 0.729 0.9 402.2 5.7 −342.8 0.1 14.6
53 0.744 6.8 426.3 1.0 −366.2 5.0 12.2
54 0.758 12.3 453.7 0.1 −393.6 6.2 14.7
55 0.772 19.7 483.6 −2.5 −424.7 7.9 16.6
56 0.786 30.9 513.0 −10.6 −456.2 10.4 16.5
57 0.801 43.5 538.7 −20.3 −484.2 12.5 16.1
58 0.815 53.3 559.3 −24.8 −506.3 12.4 18.3
59 0.829 59.1 573.0 −22.7 −521.5 10.0 23.6
60 0.844 63.1 576.4 −17.8 −527.1 6.4 29.5
(continued )
ISTUDY
Thigh Segment
Knee Hip Knee Hip
Frame Time RX RY RX RY Moment Moment
s N N N N N m N m
61 0.858 66.5 566.1 −13.2 −519.0 2.2 34.4
62 0.872 68.9 538.7 −10.1 −492.1 −2.3 37.3
63 0.887 69.7 491.1 −10.7 −442.7 −6.5 37.2
64 0.901 67.5 421.9 −16.0 −370.1 −10.1 33.6
65 0.915 60.1 335.2 −23.2 −279.2 −12.5 27.5
66 0.929 46.1 240.4 −27.5 −179.5 −13.3 20.7
67 0.944 26.4 149.3 −25.0 −82.3 −12.5 15.2
68 0.958 4.1 71.2 −16.0 2.9 −10.6 11.9
69 0.972 −16.9 6.8 −3.5 74.9 −6.8 9.3
TOR 70 0.987 −32.9 −34.3 8.9 122.0 −3.6 9.0
71 1.001 −30.4 −29.2 7.9 120.0 −4.6 10.4
72 1.015 −28.0 −23.0 10.7 113.5 −5.7 12.3
73 1.030 −25.4 −17.2 14.0 104.6 −6.4 13.6
74 1.044 −22.1 −13.4 14.8 96.0 −6.3 13.4
75 1.058 −18.1 −12.2 12.8 88.5 −5.5 11.8
76 1.072 −14.0 −12.8 9.0 80.7 −4.3 9.4
77 1.087 −10.0 −14.1 3.9 71.1 −3.0 6.8
78 1.101 −6.6 −15.6 −2.0 61.1 −1.8 4.2
79 1.115 −3.7 −17.8 −8.0 54.2 −0.6 2.0
80 1.130 −1.6 −21.5 −13.8 51.8 0.5 0.4
81 1.144 0.2 −25.9 −18.9 53.0 1.5 −0.7
82 1.158 1.7 −30.7 −22.8 55.8 2.2 −1.4
83 1.173 3.5 −35.5 −25.5 59.1 2.7 −1.7
84 1.187 6.0 −39.8 −27.3 62.0 3.1 −2.1
85 1.201 9.8 −43.0 −28.9 63.7 3.6 −3.0
86 1.215 15.1 −44.7 −30.5 65.2 4.3 −4.5
ISTUDY
87 1.230 21.7 −45.3 −31.7 68.8 5.6 −6.7
88 1.244 29.4 −45.0 −32.7 75.5 7.4 −9.4
89 1.258 37.5 −43.5 −34.2 83.8 9.7 −12.8
90 1.273 44.8 −40.3 −35.7 92.0 12.3 −16.4
91 1.287 49.6 −36.0 −34.2 99.0 14.5 −19.0
92 1.301 50.8 −31.4 −28.5 104.4 15.9 −19.7
93 1.316 48.4 −27.6 −20.9 108.1 15.9 −18.6
94 1.330 43.5 −26.5 −14.9 111.6 14.5 −15.7
95 1.344 36.8 −29.0 −12.5 115.8 11.7 −11.5
96 1.358 29.5 −33.9 −14.1 119.7 8.2 −6.7
HCR 97 1.373 23.4 −39.0 −19.4 122.3 4.9 −2.9
98 1.387 19.9 −42.8 −27.3 123.5 2.8 −1.0
99 1.401 19.4 −44.2 −36.7 121.9 2.0 −1.4
100 1.416 21.4 −42.9 −47.1 115.1 2.4 −3.9
101 1.430 24.8 −39.7 −58.3 103.2 3.2 −7.8
102 1.444 28.4 −36.1 −68.8 89.8 4.1 −12.0
103 1.459 30.3 −33.7 −75.7 80.0 4.6 −14.7
104 1.473 28.9 −33.1 −76.9 76.4 4.2 −14.9
105 1.487 24.4 −33.6 −73.9 77.2 3.2 −12.8
106 1.501 19.0 −34.0 −72.3 78.9 2.1 −10.4
ISTUDY
TABLE A.6 Segment Potential, Kinetic, and Total Energies – Foot, Leg, Thigh, and ½ HAT
Foot Segment Leg Segment Thigh Segment HAT Segment
Frame Time PE TKE RKE Total PE TKE RKE Total PE TKE RKE Total PE TKE RKE Total
s J J J J J J J J J J J J J J J J
1 0.000 1.2 1.8 0.0 3.0 9.5 8.3 0.1 17.9 36.3 12.3 0.3 48.8 203.6 18.2 0.4 222.3
2 0.014 1.3 2.3 0.0 3.6 9.6 9.2 0.1 18.9 36.2 12.2 0.3 48.8 203.8 18.3 0.3 222.4
3 0.029 1.4 2.9 0.0 4.3 9.7 9.9 0.0 19.6 36.3 11.7 0.4 48.4 204.1 17.9 0.1 222.1
4 0.043 1.4 3.5 0.0 4.9 9.8 10.2 0.0 20.0 36.4 11.1 0.4 47.9 204.6 17.2 0.1 221.8
5 0.057 1.4 4.0 0.0 5.5 9.9 10.4 0.0 20.3 36.7 10.5 0.4 47.6 205.2 16.3 0.0 221.6
6 0.072 1.4 4.6 0.0 6.0 10.0 10.4 0.1 20.5 36.9 10.0 0.4 47.4 206.0 15.6 0.0 221.7
7 0.086 1.4 5.1 0.0 6.5 10.0 10.3 0.2 20.5 37.3 9.6 0.4 47.3 207.0 15.1 0.1 222.1
8 0.100 1.3 5.5 0.1 6.9 10.1 10.2 0.3 20.5 37.6 9.3 0.3 47.2 207.9 14.8 0.1 222.8
9 0.114 1.3 6.0 0.1 7.3 10.1 10.1 0.3 20.5 38.0 9.0 0.2 47.2 208.9 14.6 0.1 223.6
10 0.129 1.2 6.3 0.1 7.6 10.1 9.9 0.4 20.4 38.3 8.6 0.2 47.1 209.8 14.4 0.2 224.3
11 0.143 1.1 6.6 0.1 7.8 10.1 9.8 0.5 20.3 38.7 8.2 0.1 47.0 210.6 14.1 0.2 224.9
12 0.157 1.0 6.9 0.1 8.0 10.0 9.6 0.5 20.2 38.9 7.7 0.1 46.7 211.2 13.9 0.3 225.4
13 0.172 0.9 7.0 0.1 8.0 9.9 9.4 0.6 20.0 39.1 7.3 0.0 46.5 211.6 13.7 0.4 225.7
14 0.186 0.8 7.1 0.1 8.0 9.9 9.2 0.6 19.7 39.3 6.8 0.0 46.1 211.9 13.7 0.4 225.9
15 0.200 0.8 7.0 0.1 7.9 9.8 9.0 0.6 19.4 39.4 6.4 0.0 45.8 212.0 13.9 0.4 226.2
16 0.215 0.7 6.9 0.1 7.7 9.7 8.6 0.6 18.9 39.4 6.1 0.0 45.5 211.9 14.4 0.3 226.6
17 0.229 0.7 6.6 0.1 7.4 9.6 8.2 0.6 18.3 39.3 5.9 0.0 45.2 211.6 15.4 0.2 227.2
18 0.243 0.7 6.2 0.1 7.0 9.5 7.6 0.6 17.7 39.2 5.7 0.0 44.9 211.1 16.7 0.2 227.9
19 0.257 0.7 5.7 0.1 6.5 9.4 7.0 0.5 16.9 39.0 5.7 0.0 44.7 210.5 18.1 0.1 228.7
20 0.272 0.8 5.0 0.1 5.8 9.3 6.4 0.5 16.1 38.7 5.8 0.0 44.6 209.7 19.6 0.1 229.4
21 0.286 0.8 4.2 0.0 5.0 9.2 5.7 0.4 15.3 38.5 5.9 0.0 44.4 208.9 20.8 0.1 229.7
22 0.300 0.8 3.3 0.0 4.2 9.1 5.1 0.3 14.4 38.2 6.1 0.0 44.3 208.1 21.5 0.0 229.7
23 0.315 0.9 2.5 0.0 3.4 9.0 4.4 0.1 13.6 37.9 6.4 0.0 44.3 207.4 21.9 0.1 229.3
24 0.329 0.9 1.7 0.0 2.6 9.0 3.8 0.1 12.8 37.7 6.7 0.0 44.4 206.7 22.0 0.1 228.8
25 0.343 0.9 1.1 0.0 2.0 8.9 3.3 0.0 12.2 37.5 7.1 0.0 44.6 206.1 22.1 0.1 228.4
26 0.357 0.9 0.7 0.0 1.5 8.8 2.8 0.0 11.6 37.4 7.4 0.0 44.8 205.7 22.5 0.1 228.3
27 0.372 0.8 0.4 0.0 1.3 8.8 2.4 0.0 11.2 37.3 7.6 0.0 44.9 205.3 23.1 0.1 228.6
28 0.386 0.8 0.3 0.0 1.1 8.7 2.1 0.1 10.8 37.2 7.7 0.0 44.9 205.1 23.9 0.1 229.2
29 0.400 0.7 0.2 0.0 1.0 8.6 1.8 0.1 10.5 37.2 7.5 0.0 44.8 205.1 24.7 0.0 229.8
30 0.415 0.7 0.2 0.0 0.9 8.6 1.6 0.1 10.3 37.3 7.3 0.0 44.5 205.3 25.2 0.0 230.5
ISTUDY
31 0.429 0.6 0.1 0.0 0.7 8.6 1.4 0.1 10.2 37.4 6.9 0.0 44.3 205.7 25.5 0.0 231.1
32 0.443 0.6 0.1 0.0 0.7 8.6 1.3 0.1 10.0 37.5 6.5 0.0 44.1 206.3 25.4 0.0 231.7
33 0.458 0.5 0.0 0.0 0.6 8.6 1.2 0.1 9.9 37.7 6.0 0.0 43.8 207.0 25.0 0.1 232.1
34 0.472 0.5 0.0 0.0 0.5 8.7 1.0 0.1 9.8 37.9 5.4 0.0 43.3 207.8 24.2 0.2 232.2
35 0.486 0.5 0.0 0.0 0.5 8.7 0.8 0.1 9.6 38.1 4.8 0.0 42.9 208.6 23.2 0.2 232.0
36 0.500 0.5 0.0 0.0 0.5 8.7 0.6 0.1 9.4 38.2 4.2 0.0 42.5 209.3 22.2 0.2 231.7
37 0.515 0.5 0.0 0.0 0.5 8.7 0.4 0.1 9.2 38.4 3.8 0.1 42.2 209.9 21.3 0.1 231.3
38 0.529 0.5 0.0 0.0 0.5 8.7 0.3 0.0 9.1 38.4 3.3 0.1 41.9 210.5 20.3 0.1 230.9
39 0.543 0.5 0.0 0.0 0.5 8.7 0.2 0.0 9.0 38.5 3.0 0.1 41.6 211.0 19.4 0.1 230.5
40 0.558 0.5 0.0 0.0 0.5 8.7 0.1 0.0 8.9 38.6 2.6 0.1 41.3 211.5 18.6 0.1 230.2
41 0.572 0.5 0.0 0.0 0.5 8.7 0.1 0.0 8.9 38.6 2.3 0.1 41.1 211.8 18.1 0.2 230.1
42 0.586 0.5 0.0 0.0 0.5 8.7 0.1 0.0 8.8 38.7 2.1 0.1 40.9 212.1 17.8 0.3 230.2
43 0.601 0.5 0.0 0.0 0.5 8.7 0.1 0.0 8.9 38.7 1.9 0.1 40.8 212.4 17.7 0.4 230.4
44 0.615 0.5 0.0 0.0 0.5 8.8 0.1 0.0 8.9 38.7 1.9 0.1 40.7 212.6 17.7 0.5 230.8
45 0.629 0.5 0.0 0.0 0.5 8.8 0.1 0.0 8.9 38.7 1.9 0.1 40.7 212.6 17.9 0.6 231.1
46 0.643 0.5 0.0 0.0 0.5 8.8 0.1 0.0 8.9 38.7 1.9 0.1 40.7 212.5 18.3 0.7 231.5
47 0.658 0.5 0.0 0.0 0.5 8.8 0.1 0.0 8.9 38.6 1.9 0.1 40.6 212.3 18.7 0.7 231.7
48 0.672 0.5 0.0 0.0 0.5 8.8 0.2 0.0 8.9 38.6 2.0 0.1 40.6 211.9 19.2 0.8 231.9
49 0.686 0.5 0.0 0.0 0.5 8.7 0.2 0.0 8.9 38.4 2.1 0.1 40.6 211.5 19.7 0.7 231.9
50 0.701 0.5 0.0 0.0 0.5 8.7 0.2 0.0 8.9 38.3 2.2 0.1 40.6 211.0 20.2 0.6 231.8
51 0.715 0.5 0.0 0.0 0.5 8.7 0.2 0.0 8.9 38.2 2.4 0.1 40.7 210.4 20.5 0.6 231.5
52 0.729 0.5 0.0 0.0 0.5 8.7 0.2 0.0 8.9 38.1 2.5 0.1 40.6 209.9 20.7 0.5 231.1
53 0.744 0.5 0.0 0.0 0.5 8.7 0.2 0.0 8.9 38.0 2.5 0.1 40.6 209.3 20.9 0.5 230.7
54 0.758 0.5 0.0 0.0 0.5 8.7 0.2 0.0 8.9 37.9 2.6 0.1 40.6 208.7 21.3 0.5 230.5
55 0.772 0.5 0.0 0.0 0.6 8.7 0.3 0.0 9.0 37.8 2.9 0.1 40.7 208.1 21.7 0.5 230.2
56 0.786 0.6 0.0 0.0 0.6 8.7 0.3 0.1 9.1 37.7 3.1 0.1 40.9 207.6 21.9 0.4 229.9
57 0.801 0.6 0.0 0.0 0.6 8.8 0.4 0.1 9.2 37.7 3.5 0.0 41.2 207.1 22.0 0.3 229.3
58 0.815 0.6 0.0 0.0 0.6 8.8 0.5 0.1 9.4 37.6 3.9 0.0 41.5 206.5 21.9 0.2 228.6
59 0.829 0.6 0.0 0.0 0.6 8.8 0.7 0.1 9.6 37.5 4.4 0.0 41.9 206.1 21.9 0.1 228.1
60 0.844 0.6 0.0 0.0 0.7 8.9 0.9 0.1 9.9 37.4 5.2 0.0 42.6 205.6 22.3 0.0 227.9
61 0.858 0.7 0.0 0.0 0.7 8.9 1.3 0.1 10.3 37.3 6.2 0.0 43.5 205.2 23.1 0.0 228.3
62 0.872 0.7 0.1 0.0 0.8 9.0 1.7 0.2 10.9 37.2 7.5 0.0 44.7 204.8 24.4 0.0 229.3
(continued )
ISTUDY
Foot Segment Leg Segment Thigh Segment HAT Segment
Frame Time PE TKE RKE Total PE TKE RKE Total PE TKE RKE Total PE TKE RKE Total
s J J J J J J J J J J J J J J J J
63 0.887 0.7 0.1 0.1 0.9 9.0 2.4 0.2 11.6 37.1 9.0 0.0 46.0 204.5 25.8 0.1 230.4
64 0.901 0.8 0.2 0.1 1.1 9.1 3.2 0.2 12.5 36.9 10.5 0.0 47.4 204.1 26.7 0.2 231.1
65 0.915 0.8 0.3 0.1 1.2 9.1 4.1 0.2 13.4 36.7 11.8 0.0 48.5 203.8 26.5 0.4 230.7
66 0.929 0.9 0.5 0.1 1.5 9.2 5.0 0.2 14.3 36.6 12.6 0.1 49.2 203.6 24.9 0.6 229.1
67 0.944 1.0 0.7 0.1 1.8 9.2 5.8 0.2 15.2 36.4 12.9 0.1 49.4 203.4 22.3 0.8 226.5
68 0.958 1.1 1.0 0.1 2.1 9.3 6.6 0.1 16.0 36.3 12.6 0.1 49.1 203.2 19.6 0.8 223.7
69 0.972 1.1 1.3 0.0 2.5 9.4 7.3 0.1 16.7 36.2 12.1 0.2 48.5 203.2 17.4 0.7 221.3
70 0.987 1.2 1.7 0.0 3.0 9.5 7.9 0.1 17.4 36.1 11.4 0.3 47.8 203.3 15.8 0.5 219.7
71 1.001 1.3 2.2 0.0 3.5 9.6 8.4 0.0 18.0 36.2 10.7 0.3 47.2 203.5 14.9 0.3 218.7
72 1.015 1.4 2.7 0.0 4.1 9.7 8.8 0.0 18.5 36.3 10.2 0.4 46.8 203.9 14.4 0.2 218.5
73 1.030 1.4 3.3 0.0 4.7 9.8 9.3 0.0 19.1 36.4 9.9 0.4 46.8 204.5 14.3 0.1 218.9
74 1.044 1.4 3.8 0.0 5.2 9.9 9.6 0.0 19.5 36.7 9.8 0.4 46.9 205.2 14.3 0.1 219.6
75 1.058 1.4 4.4 0.0 5.8 10.0 9.9 0.1 20.0 37.0 9.8 0.4 47.1 206.1 14.5 0.1 220.7
76 1.072 1.4 4.9 0.0 6.3 10.0 10.1 0.2 20.3 37.3 9.8 0.3 47.4 207.0 14.7 0.1 221.8
77 1.087 1.3 5.5 0.0 6.8 10.1 10.2 0.2 20.5 37.6 9.7 0.3 47.6 208.1 14.7 0.2 223.0
78 1.101 1.2 5.9 0.1 7.2 10.1 10.2 0.3 20.6 38.0 9.5 0.2 47.7 209.1 14.6 0.3 224.0
79 1.115 1.1 6.4 0.1 7.6 10.0 10.2 0.4 20.6 38.3 9.1 0.2 47.6 210.0 14.4 0.3 224.8
80 1.130 1.1 6.7 0.1 7.8 10.0 10.1 0.5 20.5 38.6 8.7 0.1 47.4 210.8 14.3 0.4 225.5
81 1.144 1.0 7.0 0.1 8.0 9.9 9.9 0.5 20.4 38.9 8.1 0.1 47.1 211.3 14.4 0.4 226.1
82 1.158 0.9 7.2 0.1 8.1 9.9 9.8 0.6 20.2 39.1 7.6 0.1 46.7 211.7 14.7 0.3 226.6
83 1.173 0.8 7.3 0.1 8.2 9.8 9.6 0.6 19.9 39.2 7.0 0.0 46.2 211.9 14.9 0.2 227.0
84 1.187 0.7 7.4 0.1 8.2 9.7 9.3 0.6 19.6 39.3 6.5 0.0 45.8 211.8 15.2 0.1 227.2
85 1.201 0.7 7.3 0.1 8.1 9.5 9.1 0.6 19.2 39.3 6.0 0.0 45.3 211.6 15.6 0.1 227.4
86 1.215 0.6 7.2 0.1 7.9 9.4 8.7 0.6 18.8 39.2 5.7 0.0 44.9 211.3 16.1 0.1 227.4
87 1.230 0.6 6.8 0.1 7.5 9.3 8.2 0.6 18.2 39.0 5.5 0.0 44.6 210.8 16.7 0.1 227.5
88 1.244 0.6 6.4 0.1 7.1 9.2 7.6 0.6 17.5 38.8 5.5 0.0 44.4 210.1 17.4 0.1 227.6
89 1.258 0.7 5.7 0.1 6.4 9.1 6.9 0.6 16.6 38.6 5.6 0.0 44.2 209.3 18.2 0.1 227.6
90 1.273 0.7 4.8 0.1 5.5 9.0 6.2 0.5 15.6 38.3 5.8 0.0 44.1 208.5 18.9 0.1 227.5
ISTUDY
91 1.287 0.8 3.8 0.0 4.6 8.9 5.4 0.3 14.6 38.0 6.0 0.0 44.0 207.6 19.4 0.1 227.2
92 1.301 0.8 2.8 0.0 3.6 8.8 4.6 0.2 13.6 37.7 6.4 0.0 44.0 206.7 20.0 0.2 226.8
93 1.316 0.8 1.9 0.0 2.7 8.8 4.0 0.1 12.8 37.4 6.8 0.0 44.2 205.9 20.5 0.2 226.5
94 1.330 0.8 1.2 0.0 2.0 8.7 3.5 0.0 12.2 37.2 7.4 0.0 44.6 205.1 21.0 0.3 226.4
95 1.344 0.8 0.7 0.0 1.5 8.6 3.0 0.0 11.7 37.1 7.9 0.0 45.0 204.6 21.5 0.3 226.4
96 1.358 0.8 0.5 0.0 1.3 8.6 2.7 0.0 11.3 37.0 8.4 0.0 45.4 204.2 22.3 0.3 226.8
97 1.373 0.7 0.3 0.0 1.1 8.6 2.4 0.1 11.0 37.0 8.6 0.0 45.6 204.0 23.1 0.3 227.4
98 1.387 0.7 0.2 0.0 1.0 8.5 2.2 0.1 10.8 37.0 8.6 0.0 45.6 204.0 24.0 0.2 228.2
99 1.401 0.6 0.2 0.0 0.8 8.5 2.0 0.1 10.6 37.1 8.3 0.0 45.4 204.2 24.6 0.1 228.9
100 1.416 0.6 0.1 0.0 0.7 8.5 1.7 0.1 10.4 37.2 7.8 0.0 45.1 204.7 24.8 0.0 229.5
101 1.430 0.5 0.1 0.0 0.6 8.6 1.5 0.1 10.2 37.4 7.2 0.0 44.6 205.3 24.7 0.0 230.0
102 1.444 0.5 0.0 0.0 0.6 8.6 1.2 0.1 9.9 37.6 6.4 0.0 44.0 206.1 24.2 0.0 230.3
103 1.459 0.5 0.0 0.0 0.5 8.6 0.9 0.1 9.6 37.8 5.5 0.0 43.3 206.9 23.3 0.1 230.2
104 1.473 0.5 0.0 0.0 0.5 8.6 0.6 0.1 9.3 37.9 4.6 0.0 42.5 207.6 21.8 0.1 229.5
105 1.487 0.5 0.0 0.0 0.5 8.6 0.4 0.1 9.1 38.1 3.7 0.0 41.8 208.3 19.7 0.1 228.1
106 1.501 0.5 0.0 0.0 0.5 8.6 0.2 0.1 8.9 38.2 2.9 0.1 41.1 208.9 16.9 0.1 225.9
ISTUDY
TABLE A.7 Power Generation/Absorption and Transfer – Ankle, Knee, and Hip
Muscle Power Gen(+)/ABS(−) Rate of Transfer Across Joints and Muscle Segment Angular Velocity
Leg to Foot Thigh to Leg Pelvis to Thigh
Frame Time Ankle Knee Hip Joint Muscle Joint Muscle Joint Muscle Foot Leg Thigh HAT
s W W W W W W W W W R/s R/s R/s R/s
2 0.014 −2.8 −37.3 51.4 49.4 −2.7 110.5 0.0 56.1 12.6 −3.46 −1.70 3.59 0.71
3 0.029 −0.4 −32.0 45.6 46.0 −1.2 85.6 0.0 29.6 7.2 −1.06 −0.80 3.81 0.52
4 0.043 1.3 −24.4 38.7 41.3 0.3 62.9 1.4 17.6 4.0 1.18 0.21 3.95 0.37
5 0.057 2.1 −15.9 31.6 36.8 1.4 44.7 7.2 14.8 2.4 3.14 1.24 3.98 0.28
6 0.072 2.3 −8.0 24.3 32.8 2.0 31.6 10.6 15.2 1.8 4.74 2.21 3.89 0.27
7 0.086 2.0 −2.1 16.8 28.9 2.2 22.6 11.0 14.6 1.6 5.91 3.08 3.68 0.33
8 0.100 1.6 1.0 10.0 24.6 2.1 17.4 7.6 12.2 1.4 6.66 3.81 3.36 0.41
9 0.114 1.2 1.4 5.1 20.1 2.1 14.9 3.0 8.2 1.0 7.03 4.41 2.99 0.50
10 0.129 1.0 −0.2 1.9 15.2 2.1 13.7 −0.2 2.6 0.6 7.16 4.90 2.58 0.59
11 0.143 0.8 −3.3 0.0 10.0 2.3 12.5 −2.3 −3.8 0.0 7.17 5.28 2.16 0.68
12 0.157 0.7 −7.3 −0.8 4.1 2.7 10.3 −3.3 −10.6 −0.6 7.13 5.56 1.74 0.76
13 0.172 0.7 −12.1 −0.8 −2.3 3.1 6.1 −3.6 −18.4 −1.3 7.08 5.75 1.33 0.83
14 0.186 0.7 −17.3 −0.2 −9.1 3.4 −0.8 −3.3 −27.4 −2.1 7.06 5.86 0.94 0.85
15 0.200 0.7 −22.4 0.9 −15.6 3.7 −10.0 −2.4 −36.6 −2.1 7.06 5.91 0.58 0.82
16 0.215 0.7 −27.9 2.6 −22.0 3.8 −20.9 −1.1 −45.3 −1.2 7.06 5.91 0.23 0.76
17 0.229 0.7 −34.4 5.2 −28.7 3.7 −32.6 0.0 −54.1 0.0 7.02 5.85 −0.11 0.67
18 0.243 0.7 −42.0 8.5 −36.2 3.3 −43.6 0.0 −62.3 0.0 6.90 5.73 −0.43 0.56
19 0.257 0.5 −49.9 12.2 −44.1 2.6 −52.6 0.0 −68.6 0.0 6.66 5.51 −0.70 0.46
20 0.272 0.3 −57.2 15.7 −51.3 1.6 −59.6 0.0 −73.1 0.0 6.25 5.16 −0.91 0.37
21 0.286 0.1 −62.3 18.7 −56.6 0.4 −65.4 0.0 −76.3 0.0 5.61 4.61 −1.02 0.32
22 0.300 −0.1 −62.4 21.3 −58.4 −0.5 −69.8 0.0 −76.3 0.0 4.71 3.85 −1.03 0.31
23 0.315 −0.2 −54.6 23.0 −55.6 −1.0 −72.5 0.0 −72.4 0.0 3.53 2.88 −0.94 0.34
24 0.329 −0.2 −38.4 23.2 −47.9 −0.9 −73.4 0.0 −66.8 0.0 2.11 1.75 −0.79 0.42
25 0.343 0.0 −17.7 21.1 −36.8 −0.3 −71.5 0.0 −61.7 0.0 0.49 0.58 −0.63 0.50
26 0.357 0.4 0.2 16.6 −25.2 0.2 −65.9 6.3 −58.0 0.0 −1.27 −0.51 −0.49 0.53
27 0.372 0.4 9.3 10.8 −16.1 0.4 −56.4 3.9 −55.4 0.0 −2.99 −1.39 −0.41 0.50
ISTUDY
HCR 28 0.386 4.0 53.9 44.0 −12.7 3.3 −99.0 13.3 −113.1 0.0 −4.40 −1.99 −0.39 0.41
29 0.400 −13.3 38.2 25.9 50.6 −10.7 −28.3 8.1 −47.0 0.0 −5.24 −2.33 −0.41 0.28
30 0.415 −23.5 16.0 12.3 86.6 −20.7 22.6 3.1 0.8 0.0 −5.34 −2.50 −0.41 0.12
31 0.429 −6.2 10.2 6.2 91.8 −7.4 43.2 1.6 15.1 1.3 −4.80 −2.61 −0.36 −0.06
32 0.443 −8.3 −18.6 0.4 76.0 −18.5 39.0 −2.2 −2.7 2.7 −3.88 −2.68 −0.29 −0.25
33 0.458 −0.7 −35.3 −1.3 55.5 −12.4 30.3 −3.8 −29.2 2.0 −2.85 −2.69 −0.26 −0.42
34 0.472 3.5 −55.3 −0.1 39.6 −9.3 31.1 −9.1 −43.5 0.1 −1.88 −2.60 −0.37 −0.55
35 0.486 3.3 −55.8 −0.2 29.2 −2.9 40.9 −20.3 −41.8 −2.8 −1.12 −2.41 −0.64 −0.59
36 0.500 −0.8 −38.8 −3.5 20.5 0.3 51.5 −36.2 −36.0 −4.1 −0.60 −2.15 −1.04 −0.56
37 0.515 0.3 −16.0 −9.5 11.5 −0.1 55.7 −54.9 −38.9 −4.9 −0.29 −1.88 −1.45 −0.49
38 0.529 −5.4 5.3 −6.9 3.2 0.4 52.9 −53.3 −48.4 −2.1 −0.12 −1.64 −1.81 −0.43
39 0.543 −7.3 16.6 −5.0 −2.5 0.1 46.3 −41.1 −51.6 −1.3 −0.02 −1.46 −2.05 −0.41
40 0.558 −9.0 21.0 −8.0 −6.7 0.0 37.1 −33.0 −42.9 −2.2 0.03 −1.33 −2.18 −0.47
41 0.572 −13.6 19.3 −10.4 −11.9 0.0 25.0 −25.0 −27.6 −3.7 0.06 −1.24 −2.19 −0.58
42 0.586 −12.8 18.4 −16.7 −18.5 0.0 12.9 −23.4 −12.7 −8.6 0.03 −1.18 −2.11 −0.72
43 0.601 −15.2 13.1 −14.2 −24.7 1.2 6.0 −18.3 −0.1 −10.8 −0.09 −1.16 −1.99 −0.86
44 0.615 −16.8 7.7 −9.7 −28.2 5.0 8.3 −12.8 12.1 −10.7 −0.27 −1.17 −1.87 −0.98
45 0.629 −16.6 4.7 −7.1 −28.3 9.7 18.3 −9.1 26.1 −10.8 −0.43 −1.17 −1.77 −1.07
46 0.643 −16.5 3.5 −6.3 −24.7 12.9 31.5 −7.7 41.5 −12.7 −0.51 −1.16 −1.70 −1.14
47 0.658 −18.4 3.3 −6.4 −18.7 13.7 43.7 −7.6 56.4 −16.4 −0.49 −1.15 −1.65 −1.19
48 0.672 −22.5 3.4 −7.1 −13.0 12.3 52.2 −8.0 69.0 −20.2 −0.41 −1.14 −1.63 −1.21
49 0.686 −28.1 2.9 −8.3 −10.0 10.7 54.6 −6.9 76.9 −20.9 −0.32 −1.16 −1.65 −1.18
50 0.701 −32.6 1.2 −8.7 −10.9 13.1 50.2 −2.7 78.3 −16.9 −0.34 −1.18 −1.69 −1.12
51 0.715 −31.2 −1.2 −7.8 −15.4 23.1 40.2 2.7 74.2 −11.7 −0.51 −1.20 −1.75 −1.05
52 0.729 −20.0 0.0 −11.5 −23.7 36.0 27.5 0.1 69.5 −14.6 −0.79 −1.23 −1.79 −1.00
53 0.744 −10.8 −2.5 −10.0 −36.4 58.2 14.7 6.4 68.6 −12.0 −1.09 −1.29 −1.79 −0.98
54 0.758 −3.8 −2.3 −11.7 −54.6 77.7 3.0 8.6 70.4 −14.2 −1.32 −1.39 −1.76 −0.97
55 0.772 −1.8 −1.0 −12.1 −80.0 97.4 −8.1 12.1 68.7 −15.6 −1.50 −1.53 −1.66 −0.94
56 0.786 −0.8 2.2 −10.4 −113.2 121.4 −19.1 15.7 59.9 −14.6 −1.71 −1.72 −1.51 −0.88
57 0.801 6.1 7.6 −8.5 −150.5 149.1 −27.8 16.4 48.6 −12.8 −2.01 −1.93 −1.32 −0.79
58 0.815 26.5 12.8 −8.3 −187.0 177.3 −31.0 13.8 42.8 −12.1 −2.46 −2.14 −1.11 −0.66
59 0.829 65.3 14.5 −9.8 −222.4 204.4 −28.5 9.0 44.0 −11.4 −3.10 −2.35 −0.90 −0.48
60 0.844 120.8 12.0 −11.8 −260.9 228.2 −22.8 4.3 49.0 −8.0 −3.89 −2.55 −0.67 −0.27
61 0.858 181.5 5.1 −13.0 −303.1 246.0 −15.9 0.9 54.8 −1.6 −4.76 −2.74 −0.43 −0.05
(continued )
ISTUDY
Muscle Power Gen(+)/ABS(−) Rate of Transfer Across Joints and Muscle Segment Angular Velocity
Leg to Foot Thigh to Leg Pelvis to Thigh
Frame Time Ankle Knee Hip Joint Muscle Joint Muscle Joint Muscle Foot Leg Thigh HAT
s W W W W W W W W W R/s R/s R/s R/s
62 0.872 232.2 −6.3 −12.0 −344.4 254.6 −10.7 −0.3 57.2 0.0 −5.62 −2.94 −0.14 0.18
63 0.887 263.5 −21.7 −8.1 −376.4 249.1 −12.8 0.0 49.2 7.1 −6.43 −3.13 0.19 0.41
64 0.901 272.4 −38.9 −2.6 −387.8 225.0 −24.0 0.0 27.1 18.9 −7.25 −3.28 0.56 0.64
65 0.915 254.1 −54.0 2.8 −366.0 182.6 −36.6 0.0 −3.4 23.8 −8.03 −3.36 0.97 0.86
66 0.929 203.4 −62.7 6.9 −307.3 129.2 −39.1 0.0 −28.9 22.2 −8.58 −3.33 1.40 1.07
67 0.944 130.7 −62.9 9.8 −221.6 77.0 −24.4 0.0 −37.9 18.5 −8.57 −3.18 1.86 1.22
68 0.958 60.7 −54.9 12.4 −125.8 35.5 5.8 0.0 −29.0 15.0 −7.81 −2.88 2.31 1.27
69 0.972 14.0 −35.3 14.2 −31.8 8.9 44.1 0.0 −8.1 11.0 −6.31 −2.46 2.73 1.19
TOR 70 0.987 −3.3 −18.2 18.7 40.6 −2.6 81.8 0.0 18.1 9.2 −4.28 −1.91 3.10 1.02
ISTUDY
APPENDIX B
UNITS AND DEFINITIONS RELATED

TO BIOMECHANICAL AND
ELECTROMYOGRAPHICAL MEASUREMENTS
All units used are Système International d’Unités (SI). The system is based on seven well-defined
base units and two supplementary units. Only one measurement unit is needed for any physical
quantity, whether the quantity is a base unit or a derived unit (which is the product and/or quotient
of two or more of the base units).
TABLE B.1 Base SI Units

Physical Quantity Symbol Name of SI Unit Symbol of Unit
Length l Meter m
Mass m Kilogram kg
Time t Second s
Electric current I Ampere A
Temperature T Kelvin K
Amount of substance n Mole mol
Luminous intensity I Candela cd
Plane angle θ, φ, etc. Radian rad
Solid angle Ω Steradian sr

Companion website:
ISTUDY
352 APPENDIX B
TABLE B.2 Derived SI Units

Name of
Physical Quantity Symbol SI Unit Definition
Velocity v m/s Time rate of change of position
Acceleration a m/s2 Time rate of change of velocity
Acceleration g m/s2 Acceleration of a freely falling body in a vacuum because
of gravity. At sea level g = 9.80665 m/s2
Angular velocity ω rad/s Time rate of change of orientation of a line segment in a
plane
Angular α rad/s2 Time rate of change of angular velocity
acceleration
Angular θ Radian (rad) Change in orientation of a line segment, which is given by
displacement the plane angle between initial and final orientations
Period T Second (s) Time to complete one cycle of a periodic event or, more
generally, time duration of any event or phase of an event
Frequency f Hertz (Hz) Number of repetitions of a periodic event that occurs in a
given time interval. 1 Hz equals one repetition or cycle
per second (1 Hz = 1/s)
Density ρ kg/m3 Mass per unit volume of an object or substance
Specific gravity d Ratio of the density of a substance to the density of water at
4 C
Force F Newton (N) Effect of one body on another that causes the bodies to
accelerate relative to an inertial reference frame. 1 N is
that force that, when applied to 1 kg of mass, causes it to
accelerate at 1 m/s2 in the direction of the force
application relative to the inertial reference frame
(1 N = 1 kg m/s2)
Weight G N Force exerted on a mass because of gravitational attraction;
equal to the product of the mass of the body and the
acceleration from gravity (G = m g)
Mass moment of I kg m2 Measure of a body’s resistance to accelerated angular
inertia motion about a given axis; equal to the sum of the
products of the masses of its differential elements and the
squares of their individual distances from that axis
Linear momentum p kg m/s Vector quantity possessed by a moving rigid body,
quantified by the product of its mass and the velocity of
its mass center
Angular L kg m2/s Vector of the linear momentum of a rigid body about a
momentum point; the product of the linear momentum and the
perpendicular distance of the linear momentum from the
point. For planar movement, the angular momentum is
the product of the moment of inertia in the plane about its
centroid and the angular velocity in the plane
Moment of force M N m Turning effect of a force about a point; the product of the
force and the perpendicular distance from its line of
action to that point
Pressure, normal p Pascal (Pa) Intensity of a force applied to, or distributed over, a surface;
stress, shear the force per unit area. 1 Pa is the pressure resulting from
stress 1 N applied uniformly and in a directional perpendicular
over an area of 1 m2 (1 Pa = 1 N/m2)
Linear strain ε Deformation resulting from stress measured by the
Shear strain γ percentage change in length of a line (linear strain) or the
change in
strain)
ISTUDY
APPENDIX B 353
TABLE B.2 (Continued )

Name of
Physical Quantity Symbol SI Unit Definition
Young’s modulus E Pa Ratio of stress to strain over the initial linear portion of a
Shear modulus G stress–strain curve
Work W Joule (J) Energy change over a period of time as a result of a force
acting through a displacement in the direction of the
force. 1 J is the work done when a force of 1 N is
displaced a distance of 1 m in the direction of the force
(1 J = 1 N m). Work is also the time integral of power
(1 J = 1 W s)
Mechanical energy E J Capacity of a rigid body to do work; quantified by the sum
of its potential and kinetic energies
Potential energy V J Energy of a mass or spring associated with its position or
configuration relative to a spatial reference
Gravitational potential energy of a mass m raised a distance
h above the reference level equals mgh, where g is the
acceleration from gravity. Elastic potential energy of a
linear spring with stiffness k stretched or compressed a
distance e equals k e2/2
Kinetic energy T J Energy of a mass associated with its translational and
rotational velocities. The translational kinetic energy T of
a mass m with a velocity v is 1/2 mv2. The rotational
kinetic energy T of a body rotating in a plane with a
moment of inertia I rotating with a angular velocity ω is
1/2 Iω2
Power P Watt (W) Rate at which work is done or energy is expended. The
power generated by a force is the dot product of the force
and the velocity at the point of application of the force
(P = F V). The power generated by a moment is the dot
product of the moment and the angular velocity of the
rigid body (P = M ω)
Coefficient of μ — For two objects in contact over a surface, the ratio of the
friction contact force parallel to the surface to the contact force
perpendicular to the surface
Coefficient of η N s/m2 Resistance of a substance to change in form; calculated by
viscosity the ratio of shear stress to its rate of deformation
Electrical charge q Coulomb Quantity of a negative or positive charge on any mass. The
(C) charge on an electron or proton is 1.602 × 10−19 C, or 1
C, has the charge of 6.242 × 1018 electrons or protons (1
A [ampere] = 1 C/s)
Voltage, electrical E Volt (V) Potential for an electrical charge to do work (1 V = 1 J/C)
potential
Electrical R Ohm (Ω) Property of a conducting element that opposes the flow of
resistance electrical charge in response to an applied voltage (1 Ω =
1 V/A)
Electrical C Farad (F) Property of an electrical element that quantifies its ability to
capacitance store electrical charge. A capacitance of 1 F means that 1
C of charge is stored with a voltage change of 1 V (1
F = 1 C/V)
ISTUDY
354 APPENDIX B
NOTES TO TABLES B.1 AND B.2
1. Prefixes are used to designate multiples or submultiples of units.
Prefix Multiplier Symbol Examples

6
Mega 10 M Megahertz (MHz)
Kilo 103 k Kilowatt (kW)
Centi 10−2 c Centimeter (cm)
Milli 10−3 m Millisecond (ms)
Micro 10−6 μ Microvolt (μV)
2. (a) When a compound unit is formed by multiplication of two or more units, the symbol for
the compound unit can be indicated as follows:
N m or N m, but not Nm
(b) When a compound unit is formed by dividing one unit by another, the symbol for the
compound unit can be indicated as follows:
kg m3 or as a produce of kg and m − 3 , kg m − 3
3. Because the symbol for second is s, not sec, the pluralization of symbols should not be
done; for example, kgs may be mistaken for kg s, or cms may be mistaken for cm s.
ISTUDY
INDEX
Accelerometers, 38–39 moment of inertia, 73–74, 78, 80–81, 109, 128, 239
Anthropometry, 67–86 ponderal index, 68
body, 67–86
experimental measures, 78–82 Cardan rotation sequences, 146–148
muscle, 82–84 Center of pressure, 25, 77–78, 90, 107–109, 239
problems, 86, 87 Cocontraction, 125–127, 136, 241–242
tables, 71, 76, 77, 83, 84 Degrees of freedom, 41, 151, 218, 274
Balance board, 79 Efficiency:
Biomechanical: mechanical, 123–124
analyses, 4, 5 muscle, 123–124
assessments, 2, 3 negative, 124
measurements, 2, 3, 36–42 positive, 124
models–inverted pendulum, 108–110, 241–242 work, 124
monitoring, 3, 4 Electromyogram:
motor synergies, 7, 236 biophysical basis, 189–195
gait initiation, 243–244 common mode rejection, 196, 199–202
gait termination, 244–246 cross-talk, 17, 192, 202–204
load/unload mechanism, 239–240 filters, 199, 205–208
role of nervous system, 274–276 frequency content, 194–199, 204, 205, 207, 211, 212
support moment, 107, 236–237 gain, 196
trunk balance in gait, 239–240 input impedance, 196, 197
Body: integrated, 207
center of gravity, 69, 107 linear envelope, 205–208
center of mass, 69–73, 75–79, 107–109, 134, 216, motor unit action potential, 189–195
226, 236, 239–240 in muscle model, 227–229
density, 68–70 presence of hum, 16, 17, 195, 199, 259
energy, 73, 115–117, 127–129, 132–133 recording amplifiers, 195–204

Companion website:
ISTUDY
356 INDEX
Energy: Glove transducer, 40

absorption, 114, 116–117, 119, 126–128, Goniometers, 36–37
134–135, 137–139, 142–143, 168–169
exchange, 128–130, 132, 134, 137 Heats of contraction, 127
generation, 106, 114, 116–117, 126–127, Hill-type muscle model, 220, 227
134–135, 137–139, 142–143, 168–169, 190,
207, 219, 249, 290–291 Imaging techniques, 134, 253, 257
kinetic, 4–5, 7–8, 23, 29, 32, 62, 65, 67–69, 75, 78, cinematography, 43
88, 90, 114–116, 128–134, 141, 143, lens optics, 42
145–146, 148, 150, 152–154, 156, 158, optoelectric, 27, 29, 42, 47, 51
161–170, 216, 236, 244, 246, 284 television, 43–44, 57, 65–66, 134
law of conservation of, 116, 136, 139, 216 Inefficiency, 123–124, 126
metabolic, 6, 117, 123–124, 126–127, 133, 142, Inverse solution, 4, 89–90, 99, 110, 136, 263–264
174, 176, 211
potential, 22, 49–51, 65–66, 83, 90, 102, 115–116, Kinematics, 8, 33–34, 36, 38–42, 44, 46, 48–52, 54,
128–134, 141, 172–175, 177, 189–195, 205, 56, 58, 60, 62, 64, 66, 76, 82, 89, 93, 100,
208, 212, 214, 216–217, 222–223, 247, 102, 104, 108, 152–153, 213, 215, 217, 222,
249–250, 255–256, 259, 261, 263–268, 224–225, 228, 231, 236, 258, 274, 281,
270–271, 273–274, 276, 279, 282, 286 283–284, 286, 288, 293
problems, 141–143 acceleration calculation, 54
spring, 180, 184–185, 208, 219, 291 angles, 4, 21, 23, 29–31, 35, 37–38, 40, 47,
storage, 23, 80, 116, 128–129, 131, 180, 233 62–63, 67, 83, 98, 114, 150, 154, 158, 161,
total body, 25, 36, 69, 72–73, 75–79, 86, 91, 109, 192, 209, 218, 220, 222, 224–225, 228, 239,
117, 123, 128, 132–133, 235–236, 239, 242, 276, 287–288, 291, 293
244–245 axes of rotation, 81
transfer, 4, 14, 122–123, 132, 137–139, 141–143, conventions, 35, 62
196, 207, 226, 276, 280, 293 direct measures, 75
Euler rotation, 158, 161 imaging techniques, 134, 253, 257
problems, 64–65
Finite differences, 57, 60 three-dimensional, 36, 76, 99, 145, 148,
Foot pressure measurement, 104 162–163, 170, 181, 183–185, 216, 233,
Force platforms, 49, 100, 239–240, 275 272, 286, 288
Forces: two-dimensional, 46, 145, 185, 205
bone-on-bone, 7, 91–93, 110–111, 113 variables, 4–5, 8–9, 29, 34–36, 62–63, 75, 89, 110,
gravitational, 39, 70, 90, 93, 96, 108, 118, 121, 139, 145–146, 148, 150, 152, 154, 156,
128–129, 145, 162, 166, 170, 184–185, 158, 161–164, 166, 168, 170, 187, 189, 192,
187, 235 208–209, 211, 221, 226–227, 230–231, 244,
ground reaction, 5, 49, 90, 94–95, 99, 102–103, 246, 263, 276, 289
105, 107, 112, 114, 118, 162, 185, 215, 217, velocity calculation, 54
225–226, 237–240, 243, 284, 289 Kinesiology, 1, 40, 203–205, 247
inertial, 17, 24, 39–40, 66, 68–69, 73, 75, 77, 87, Kinetics, 89–90, 92, 94, 96, 98, 100, 102, 104, 106,
95, 153, 167, 184, 224, 235, 258 108, 110, 112, 114, 150, 153, 162, 170, 188,
muscle, 4–7, 14, 17–18, 32, 37, 47, 50, 68–69, 215, 282, 285–286, 293
82–97, 100, 104–105, 109, 111, 113–116, problems, 111–112
118–120, 123–128, 132–133, 135–137, 139,
142–143, 167, 170–195, 197–199, 202–215,
Link-segment model, 4, 6, 89–92, 116, 215–217
217–234, 236, 239, 242, 246–253, 260,
assumptions, 55, 90, 104, 110, 116,
265–266, 269–271, 274–277, 279–280, 284,
216–217, 223
286, 293
reaction, 4–5, 8, 23–24, 38, 49, 79, 89–97,
99–105, 107, 109–110, 112–114, 118, 123, Moments-of-force, 8
136, 139–140, 162–165, 167, 169, 185, interpretation, 1–2, 5, 34, 58, 101, 105, 118, 124,
215–217, 225–226, 237–240, 243, 127, 132, 166, 168–170, 274, 280
284, 289 Motor unit, 6, 171, 173, 176–177, 187–195, 209,
Forward solution, 215–217, 263–264 211–212, 247, 264
assumptions, 216–217 action potential, 172–175, 177, 189–192,
examples, 217 194–195, 208, 212, 249–250, 271
formulation, 191, 222 final common pathway, 6, 189, 247
review, 216–217 motor end plate, 171, 173, 189
ISTUDY
INDEX 357
muscle twitch, 4, 175–176, 207, 271 Signal Processing:

recruitment, 6, 172–174, 176–177, 188–189, 207, aliasing, 23
211, 215, 222, 233, 264–266 analog-to-digital conversion, 19, 22
size principle, 6, 173–174, 176–177 correlation analyses, 8–9, 11, 13, 15, 17
Muscle: autocorrelation, 8, 10–13, 16–17, 28, 194, 211
active state, 187 cross-correlation, 8–11, 13–15, 17–18, 44,
concentric contraction, 118, 221 203–204, 240
contractile element, 171–172, 177–181, 184, 190, covariance analysis, 30, 236
211, 227 ensemble averaging, 8, 29, 31
cross-sectional area, 4, 67, 69, 82–84, 86, 88, 111, Fourier, harmonic, 14–15, 19–22
172, 220, 284 Fourier, white noise, 28–30
eccentric contraction, 118 Pearson, 8–11, 246
fatigue, 3, 123, 173, 189, 194, 208, 211–213, 220, record length, 25
223, 276–277 sampling theorem, 22–23
fibers, 82–84, 171–175, 183, 189–190, 192, 195, smoothing of data
198, 202, 204–205, 210–212, 220–221, 227, curve fitting, 54–55
247, 277 harmonic reconstitution, 28
force-length characteristics, 177–179, 181, Size principle, 6, 173–174, 176–177
184–185 Spatial reference systems, 35
force-velocity characteristics, 181–183, 185 global reference system, 145
mass, 83 local, anatomical reference system, 145–146
modeling, 2, 89, 170, 215–224, 226–232, 234 marker axis system, 145–146, 148, 150
moment arm length, 85, 113 three-dimensional, 36, 76, 99, 145, 148, 162–163,
origin and insertion, 67, 82, 84, 111, 219–220 170, 181, 183–185, 216, 233, 272, 286, 288
parallel elastic element, 178–179, 184 two-dimensional, 46, 145, 185, 205
pennation, 83, 220, 227
sarcomere, 172, 177, 188, 220 Work:
series elastic element, 178–180 examples, 121–122
stress, 3, 83–84, 111, 279, 281–284, 289, external, 5, 34, 51, 81, 89–90, 98, 102–103,
292–293 116–120, 124, 127–128, 133–137, 143, 161,
tremor, 211 168–169, 179, 183, 215, 217, 231, 235–236,
twitch-type, 174–176, 220–221 258, 265, 272, 277, 289–290, 292–293
against gravity, 126–128, 184
Power, mechanical internal, 34, 77, 81, 111, 114, 116–118, 124, 127,
balance of, 17, 69, 98 133, 135–137, 161, 166, 169–170, 179–184,
muscle, 119–120, 135–139 196–197, 215–218, 226, 235, 239, 250–251,
283–285, 289–290, 292
Recruitment, 6, 172–174, 176–177, 188–189, 207, muscle, 4–7, 14, 17–18, 32, 37, 47, 50, 68–69,
211, 215, 222, 233, 264–266 82–97, 100, 104–105, 109–111, 113–116,
118–120, 123–128, 132–133, 135–137, 139,
142–143, 167, 170–195, 197–199, 202–215,
Segment: 217–234, 236, 239, 242, 246–253, 260,
definitions, 47, 88, 170 265–266, 269–271, 274–277, 279–280, 284,
density, 13–14, 20, 68–71, 73, 75, 77, 83, 86, 150, 286, 293
191, 198, 211, 259–264 negative, 5, 9, 14, 18, 23, 31, 63, 85, 94–95, 100,
energy, 116, 139 106–107, 117–120, 122–128, 133–134,
lengths, 67–68, 75, 83, 85, 178, 185, 223 136–137, 139, 143, 152, 171, 183, 185–186,
mass, 69, 75–77, 89–90 192, 210–211, 240, 249, 258, 262, 274, 277
mass center, 74–75, 90, 224–226 positive, 5, 7, 9–10, 31, 35–36, 62–63, 85, 94–97,
moment of inertia, 73–74, 78, 80–81, 86–87, 90, 100, 106–107, 117–120, 123–128, 132–137,
96, 105, 109, 128, 239 139, 152, 168, 171, 183, 186, 192, 206,
radius of gyration, 71, 73–75, 78, 86 210–211, 241, 258
ISTUDY
WILEY END USER LICENSE AGREEMENT
Go to www.wiley.com/go/eula to access Wiley’s ebook EULA.
ISTUDY

Stephen J. Thomas, Joseph A. Zeni, David A. Winter - Winter's Biomechanics and Motor Control of Human Movement-Wiley (2022)

Uploaded by

Copyright:

Available Formats

Stephen J. Thomas, Joseph A. Zeni, David A. Winter - Winter's Biomechanics and Motor Control of Human Movement-Wiley (2022)

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Stephen J. Thomas, Joseph A. Zeni, David A. Winter - Winter's Biomechanics and Motor Control of Human Movement-Wiley (2022)

Uploaded by

Copyright:

Available Formats

ISTUDY

DAVID A. WINTER (DECEASED)

Limit of Liability/Disclaimer of Warranty

Library of Congress Cataloging-in-Publication Data

Cover Image: © John Emerson

Set in 10.75/12pt Times by Straive, Pondicherry, India.

STEPHEN J. THOMAS, PHD, ATC

2.1.7 Application of Autocorrelations, 16

3.0 Historical Development and Complexity of Problem, 34

3.6.3 Velocities – Linear and Angular, 63

5 Kinetics: Forces and Moments of Force 89

5.3 Bone-on-bone Forces During Dynamic Conditions, 110

6.0 Introduction, 115

7 Understanding 3D Kinematic and Kinetic Variables 145

7.0 Introduction, 145

7.3 Determination of Segment Angular Velocities and Accelerations, 158

8 Muscle Mechanics 171

9 Kinesiological Electromyography 189

9.0 Introduction, 189

9.2 Recording of the Electromyogram, 195

10 Modeling of Human Movement 215

10.0 Introduction, 215

11 Static and Dynamic Balance 235

11.0 Introduction, 235

12 Central Nervous System’s Role in Biomechanics 247

13 A Case-Based Approach to Interpreting Biomechanical Data 281

A. Kinematic, Kinetic, and Energy Data 295

B. Units and Definitions Related to Biomechanical and Electromyographical

Thomas Hulcher, M.S., Biomechanist, Department of Physical Therapy, Jefferson College of

Brian A. Knarr, PhD, Associate Professor, Department of Biomechanics, College of Education,

Todd J. Leutzinger, PhD, Instructor, Department of Biomechanics, College of Education,

Kristen F. Nicholson, PhD, Assistant Professor, Department of Orthopaedic Surgery, Wake

David A. Winters (Deceased), University of Waterloo, Waterloo, Ontario, Canada

S TEPHEN J. T HOMAS , P H D, ATC

S TEPHEN J. T HOMAS , P H D, ATC

This book is accompanied by a companion website

The website includes:

1.0.1 Importance of Human Movement Analysis

Winter’s Biomechanics and Motor Control of Human Movement, Fifth Edition.

1.0.2 The Interprofessional Team

1.1 MEASUREMENT, DESCRIPTION, ANALYSIS, AND ASSESSMENT

1.1.1 Measurement, Description, and Monitoring

1.1.3 Assessment and Interpretation

1.2 BIOMECHANICS AND ITS RELATIONSHIP WITH PHYSIOLOGY

Because biomechanics is still an expanding discipline, it is important to identify its interaction

i.e, Support moment, Ms = Mα + Mk + Mh (Final common mechanical pathway)

2.1 AUTO- AND CROSS-CORRELATION ANALYSES

Winter’s Biomechanics and Motor Control of Human Movement, Fifth Edition.

2.1.1 Similarity to the Pearson Correlation

Figure 2.1 Scatter diagram of variable x against variable y

2.1.2 Formulae for Auto- and Cross-Correlation Coefficients

where x(t) has zero mean.

where x(t) and y(t) have zero means.

2.1.3 Four Properties of the Autocorrelation Function

Rxx 0 + Rxx 0 − 2Rxx τ ≥ 0

Substituting t = (t − τ) and taking the derivative, we have dt = dt :

Since T is one period of sin(ωt), ∴ sin(2ωT + ωτ) − sin(ωτ) = 0 for all τ.

∴ Rxx τ = Rss τ + Rnn τ (2.7)

2.1.4 Three Properties of the Cross-Correlation Function