Biomechanics and Motor Control of Human Movement - Ch3
Biomechanics and Motor Control of Human Movement - Ch3
Biomechanics and Motor Control of Human Movement - Ch3
KINEMATICS
David A. Winter
45
46
KINEMATICS
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 bodys center of mass. On the other hand, a mechanical power analysis
of an amputees gait may require almost all the kinematic variables that are
available.
3.1
KINEMATIC CONVENTIONS
In order to keep track of all the kinematic variables, it is important to establish a convention system. In the anatomical literature, a definite convention
has been established, 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, that is, they 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. However, when instruments are attached to the body, the data
become relative, and we lose information about gravity and the direction of
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 is Y , the direction of progression (anteriorposterior) is X , and the
sideways direction (mediallateral) is Z . 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 XY plane are measured from 0 in
the X direction, with positive angles being counterclockwise. Similarly, in
the YZ plane, angles start at 0 in the Y direction and increase positively
counterclockwise. The convention for velocities and accelerations follows
correctly if we maintain the spatial coordinate convention:
x
y
z
x
y
z
The same applies to angular velocities and angular accelerations. A counterclockwise angular increase is a positive angular velocity, . When is
increasing, we calculate a positive angular acceleration, .
Figure 3.1
47
An example taken from the data on a human subject during walking will
illustrate the convention. The kinematics of the right leg segment (as viewed
from the right side) and its center of mass were analyzed as follows:
= 2.34 rad/s,
= 14.29 rad/s2 ,
ax = 9.27 m/s2 ,
y = 0.021 m/s,
x = 0.783 m/s
ay = 0.31 m/s2
This means that the leg segment is rotating clockwise but is decelerating
(accelerating in a counterclockwise direction). The velocity of the legs center
of mass is forward and very slightly upward, but it is decelerating in the
forward direction and accelerating downward.
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 changing with time:
1. Position (x, y, z) of segment center of mass
2. Linear velocity (x , y , z ) of segment center of mass
48
KINEMATICS
3.
4.
5.
6.
Linear acceleration (x , y,
z ) of segment center of mass
Angle of segment in two planes, xy , yz
Angular velocity of segment in two planes, xy , yz
Angular acceleration of segment in two planes, xy , yz
Note that the third angle data are redundant; any segments 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 = 12 segments), movement in three-dimensional (3D)
space required 15 12 = 180 data variables. It is no small wonder that
we have yet to describe, let alone analyze, some of the more complex
movements. Certain simplifications can certainly reduce the number of
variables to a manageable number. In symmetrical level walking, for
example, we can assume sagittal plane movement and can normally ignore
the arm movement. The head, arms, and trunk (HAT) are often considered
to be a single segment, and assuming symmetry, we need to collect data
from one lower limb only. The data variables in this case (four segments,
one plane) can be reduced to a more manageable 36.
3.2
3.2.1 Goniometers
A goniometer is a special name given to the electrical potentiometer that can
be attached to measure 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 . Thus, the voltage on the wiper
arm is = kE = k1 volts. Note that a voltage proportional to requires a
potentiometer whose resistance varies linearly with . 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.
49
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.)
50
KINEMATICS
Disadvantages
1. Relative angular data are given, not absolute angles, thus severely limiting the datas 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.
3.2.2 Special Joint Angle Measuring Systems
More recently in the area of ergonomics, a special glove system has been
developed to measure the kinematics of the fingers and the thumb. Figure 3.4
shows the construction of the 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. Each transducer is a loop of fiber-optic cable
with a constant infrared source and is etched in the region of the joint of
interest. As the joint flexes, the fiber bends and light escapes; the greater
the bend, the more light that escapes. Thus, the flexion angular displacement
is detected as a reduction in light intensity received by the detector and is
precalibrated against the bending angle. A major use for such a system has
been in the study of repetitive strain injuries (cf. Moore et al., 1991).
3.2.3 Accelerometers
As indicated by its name, an accelerometer is a device that measures acceleration. Most accelerometers are nothing more than force transducers designed
to measure the reaction forces associated 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. The acceleration can be toward or away from the face of
the transducer; the latter is indicated by a reversal in sign of the signal. In
most movements, there is no guarantee that the acceleration vector will act
at right angles to the face of the force transducer. The more likely situation
is depicted in Figure 3.5, with the acceleration vector having a component
normal to the transducer and another component tangent to the transducer
face. Thus, the accelerometer measures the an component. Nothing is known
51
Figure 3.4 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.)
52
KINEMATICS
Figure 3.6 Two movement situations where the acceleration in space is identical but
the normal components are quite different.
53
R1
+
Supply
voltage
+
V
Unbalanced
bridge
Ra
1
m
Balance
potentiometer
Rb
a
R1
Supply
voltage
Figure 3.7 Electrical bridge circuit used in most force transducers and accelerometers.
See text for detailed operation.
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. Cost of accelerometers can be excessive if a large number are used;
also the cost of the recorder or analog-to-digital converter may be high.
3. If a large number are used, they can encumber movement.
4. Many types of accelerometers are quite sensitive to shock and are easily
broken.
5. The mass of the accelerometer may result in a movement artifact, especially in rapid movements or movements involving impacts.
3.3
The Chinese proverb A picture is worth more than ten thousand words holds
an important message for any human observer, including the biomechanics
researcher interested in human movement. Because of the complexity of most
movements, the only system that can possibly capture all the data is an imaging system. 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 be limited to three different types: movie camera, television, and
54
KINEMATICS
Figure 3.8
image.
Simple focusing lens system showing relationship between the object and
(3.1)
The imaging systems used for movement studies are such that the object
lens distance is quite large compared with the lensimage distance.
Therefore,
1
0,
u
1
1
= ,
f
v
or f = v
(3.2)
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 maximum accuracy, it is highly desirable that the image be as
large as possible. Thus, it is advantageous 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).
55
Figure 3.9 Differences in the focal length of wide angle, normal, and telephoto lenses
result in an image of the same size.
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 Cinematography
Many different sizes of movie cameras are available; 8-mm cameras are the
smallest. (They actually use 16-mm film, which is run through the camera
twice, then split into two 8-mm strips after it is developed.) Then there are
16 mm, 35 mm, and 70 mm. The image size of 8 mm is somewhat small
for accurate measurements, while 35-mm and 70-mm movie cameras are
too expensive to buy and operate. Thus, 16-mm cameras have evolved as
a reasonable compromise, and most high-speed movie cameras are 16 mm.
There are several types of 16-mm cameras available. Some are spring driven;
others are motor driven by either batteries or power supplies from alternating
current sources. Battery-driven types have the advantage of being portable to
sites where power is not available.
56
KINEMATICS
The type of film required depends on the lighting available. The ASA rating is a measure of the speed of the film; the higher the rating, the less light
is required to get the same exposure. 4-X reversal film with an ASA rating of
400 is a common type. Higher ASA ratings are also available and are good
for a qualitative assessment of movement, especially faster-moving sporting events. However, the coarse grain of these higher ASA films introduces
inaccuracies in quantitative analyses.
The final factor that influences the lighting required is the shutter speed
of the camera. The higher the frame rate, the less time that is available to
expose film. Most high-speed cameras have rotating shutters that open once
per revolution for a period of time to expose a new frame of unexposed
film. The arc of the opening, as depicted in Figure 3.10, and the speed of
rotation of the shutter decide the exposure time. For example, at 60 frames
per second, using a 3 factor shutter, the exposure time is 1/180 s. The amount
of light entering will be the same as a normal (still) camera set to a speed
of 1/180 s.
To make the final settings, we use an exposure meter to measure the light
intensity on the human subject. For a given filming, the variables that are
preset are film ASA, shutter factor, and frame rate. The frame rate is set
low enough to capture the desired event but not so high as to require extra
lighting or result in film wastage. To understand the problem associated with
the selection of an optimal rate, the student is referred back to Section 2.2.4.2
on the sampling theorem. The final variable to decide is the f -stop. The light
meter gives an electrical meter reading proportional to the light intensity,
such that when the film ASA and exposure time are set, the correct f -stop
Figure 3.10 Various-factor shutters used in movie cameras. Film is exposed during
the opening arc and is advanced while the shutter in closed.
57
can be determined. Thus, with the movie camera set at the right frame rate,
f -stop, and range, the filming is ready to commence.
Figure 3.11 Typical arrangement for the microcomputer digitization of data coordinates from movie film. Foot pedal allows operator to transfer coordinate data to the
computer at a rate of about 10 coordinate pairs per minute. Digitizing error is about
1 mm rms with the camera located 4 m from the subject.
58
KINEMATICS
The human error involved in this digitizing has been found to be random
and quite small. For a camera 4 m from a subject, the root-mean-square (rms)
noise present in the converted data has been measured at 11.5 mm.
3.3.4 Television
The major difference between television and cinematography is the fact that
television has a fixed frame rate. The name given to each television image
is a field. In North America, there are 60 fields per second; in Europe, the
standard is 50 fields per second. Thus, television has a high enough field rate
for most movements, but it is probably too low for a quantitative analysis
of rapid athletic events. The f -stop, focus, and lighting for television can be
adjusted by watching the television monitor as the controls are varied. Many
television cameras have electronic as well as optical controls that influence
brightness and contrast, and some have built-in strobe lighting. Also, focus
can be adjusted electronically as well as optically. The major advantage of
television is the capability for instant replay, which serves both as a quality
control check and as an initial qualitative assessment. Second, the television
signal can be digitally converted by a frame grabber for immediate analysis.
3.3.4.1 Television Imaging Cameras. Some technical problems can result
from the use of standard vidicon television cameras. The strong signal from
a reflective marker produces a distinct circular image when the marker is
not moving. However, when there is a rapid marker movement, the circular
image blurs and produces a trailing edge. Thus, the triggering threshold for
conversion into two levels (black and white) must be carefully set to ensure
a circular digitized image. A more reliable way to get rid of the blur is to
use a strobe system, which results in the exposure of the TV imaging tube
for a millisecond or less. The strobe, in effect, acts as an electronic shutter.
Strobe systems also eliminate a second problem associated with a continuously exposed imaging tube: skewing of the marker coordinates because of
the time delay in the scanning from the top of the image to the bottom of the
image. It takes about 15 ms to scan one TV field; thus, a head marker could
be scanned 10 ms before a foot marker. The strobe system freezes all marker
images at the same point in time in the same way that a movie camera does.
Newer charge coupled diode (CCD) cameras have mechanical or electronic
shutter controls that eliminate both blurring and skewing. A further development is the infrared camera, which does not use visible light and is not
influenced by reflections from light sources other than those sources required
to get the desired circular reflection from the markers. Figure 3.12 shows a
typical infrared camera mounted permanently from the ceiling in a clinical
gait laboratory. The active infrared lights form a donut shape about the
camera lens and are pulsed at 120 Hz for a period of less than a millisecond.
59
Figure 3.12 Typical infrared television camera mounted permanently in a clinical gait
laboratory. The infrared lights form a donut shape around the lens and are pulsed for
a short time each frame to freeze the image prior to scanning. (Courtesy of the Gait
Analysis Laboratory, Connecticut Childrens Medical Center, Hartford, CT.)
This camera is one of six to twelve cameras that could be mounted around
the gait laboratory. Thus, the reflected infrared light from the markers is the
only light that is picked up by the camera, and since it is a pulsed source,
the marker images are frozen in time. Figure 3.13 depicts such an arrangement in a clinical gait laboratory along with the spherical reflective markers
mounted on a young patient.
60
KINEMATICS
Figure 3.13 Gait assessment of a cerebral palsy patient 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. (Courtesy of the Gait Analysis
Laboratory, Connecticut Childrens Medical Center, Hartford, CT.)
61
62
KINEMATICS
Figure 3.14 An OPTOTRAK system with three lenses, each with a linear diode array.
The outside two lenses face slightly inward and each defines a vertical plane, while the
middle lens defines a horizontal plane. See Figure 3.15 to see how these three diode
arrays define a marker in 3D space. (Courtesy of the Gait and Posture Laboratory,
Department of Kinesiology, University of Waterloo, Waterloo, Ont., Canada.)
63
Figure 3.15 OPTOTRAK system with two outside lenses facing slightly inward with
their diode arrays, each defining markers in a vertical plane. Any marker on the intersection of these two planes will define all markers on this vertical line. The middle
lens array defines all markers in a horizontal plane. Thus, the intersection of a marker
on this horizontal plane with the vertical line will define the unique coordinates of a
marker in 3D space.
the number and location of markers is changing from day to day, there are
no problems with marker labeling, which makes this system flexible for the
changing research requirements. Also, because of the precision of the IRED
array, the precision of the x , y, z coordinates is better than that of TV systems,
which are constrained by the distance between the scan lines. The precision
for an OPTOTRAK camera mounted as shown in Figure 3.15 at a distance of
4 m from the subject is 0.03 mm with noise = 0.015 mm (Gage et al., 2004).
Possible disadvantages are the number of IREDs that can be mounted and
the potential encumbrance of the cables connecting the power source to the
active IREDs.
64
KINEMATICS
Disadvantages
1. Most multiple-camera systems are expensive (cine, TV, optoelectric), as are the digitizing and conversion systems for all imaging
sources.
2. For film, the turnaround time for development may be a problem, and
the labor to digitize film coordinates may also be a constraint. The
digitizing errors, however, are less than those from many commercial
imaging systems.
3. Encumbrance and time to fit wired light sources (e.g., IREDs) can be
prohibitive in certain movements, and the number of light sources is
limited.
4. Some imaging systems (e.g., IREDs) cannot be used outside in daylight.
3.4
65
errors in the converted data. It is, therefore, essential that the raw data be
smoothed, and in order to understand the techniques used to smooth the data,
an appreciation of harmonic (or frequency) analysis is necessary. The theory
of harmonic analyses has been covered in Section 2.2; however, there are
some additional special problems with the processing of kinematic data that
are now discussed.
3.4.2 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 Y (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 Hz, 3 Hz, 4 Hz, and so on. Normal walking has been
analyzed by digital computer, 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 3.16, 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
1.00
mean
2s
.10
.01
.001
0
9 10 11 12 13 14 15 16 17 18 19 20
Harmonic Number
Figure 3.16 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 from the Journal of Biomechanics.)
66
KINEMATICS
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 3.16. Here you can see evidence of
higher-frequency components extending up to the 20th harmonic, which was
the highest frequency analyzed.
3.4.3 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 differentiation of a signal containing
additive higher-frequency noise. Suppose that the signal can be represented
by a summation of N harmonics:
x=
N
n=1
Xn sin (n0 t + n )
(3.3)
67
(3.4)
n=1
Ax
(3.5)
n=1
Thus, the amplitude of each of the harmonics increases with its harmonic
number; for velocities they increase linearly, and for accelerations the increase
is proportional to the square of the harmonic number. This phenomenon
is demonstrated in Figure 3.17, where the fundamental, second, 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 (velocity) 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.
In the trajectory data for gait, x1 might be 5 cm and x20 = 0.5 mm. The 20th
harmonic noise is hardly perceptible in the displacement plot. In the velocity
calculation, the 20th harmonic increases 20-fold so that it is now one-fifth
that of the fundamental. In the acceleration calculation, the 20th 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 3.19, which
plots the acceleration of the toe during walking. The random-looking signal is
the raw data differentiated 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.
68
KINEMATICS
(3.6)
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 n , as was discussed previously in Section 2.2.4.5.
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 improper selection of the inflection points. These inflection points
must be determined from the noisy data and, thus, are strongly influenced by
the very noise that we are trying to eliminate.
3.4.4.2 Digital FilteringRefiltering 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 Section 2.2.4.4. 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:
c =
(tan(fc /fs ))
C
(3.8)
K=
69
70
KINEMATICS
finite differences from the filtered data, is plotted in Figure 3.19. 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.
3.4.4.3 Choice of Cutoff Frequency Residual Analysis. There are several ways to choose the best cutoff frequency. The first is to carry out a
harmonic analysis as depicted in Figure 3.16. 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 method is to do a
residual analysis 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 3.20 shows a theoretical plot of residual versus frequency.
The residual at any cutoff frequency is calculated as follows [see Equation
(3.9)] for a signal of N sample points in time:
N
1
(Xi X i )2
(3.9)
R(fc ) =
N
i =1
71
Toe Marker
3200
1600
1600
3200
0.0
1.0
1.5
Time (s)
2.0
2.5
3.0
Figure 3.19 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.)
Figure 3.20 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.
72
KINEMATICS
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 balance 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 fc1 , 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 3.21 is a plot of the residual of four markers from one stride of gait
data, and both vertical and horizontal coordinates were analyzed (Wells and
Winter, 1980). As can be seen, the straight regression line that represents the
noise is essentially the same for both coordinates on all markers. This tells
25
Motion
horizontal vertical
Marker
20
Rib
Hip
15
Heel
Ball
Best fit linear regression for harmonics 11 to 20:
10
0
0
4
2
6
4
8
6
10
12
harmonic n
8
14
16
10
12
Frequency Hz
14
18
16
20
18
Figure 3.21 Plot of the residual of four markers from a walking trial; both vertical
and horizontal displacement data. Data were digitized from movie film with the camera
5 m from the subject.
73
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 cine
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.
3.4.4.4 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
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, velocities, or accelerations were
being considered. Unfortunately, their reference displacement signal was
reconstituted from a harmonic analysis, and in Section 2.2.4.5 this technique
was shown to have major problems because of 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 suprising when we consider that the acceleration increases as
the square of the frequency (Section 3.4.3); thus, 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, t = 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:
fc,2 = 0.06 fs 0.000022 fs2 + 5.95/
(3.10)
where fs is the sampling frequency and 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 shows a reasonable
match between the accelerometer data and the filtered film data, except that
the lag of the filtered data suggests that a second-order low-pass filter was
used rather than the desired fourth-order zero-lag filter.
74
KINEMATICS
(3.11)
Note that 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:
(t) = 70 + 1260t 5160t 2 + 8000t 3 5100t 4
+ 1050t 5 + 123t 6 29.5t 7 rad/s2
(3.12)
75
(a) 2.40
Theta vs time
Analog
Raw film data
1.60
4.00
1.20
0.80
2.00
0
2.00
0.40
0
1.20
Alpha vs time
Analog
Filtered finite
6.00
Alpha(rad/s/s), 10
Theta, rad
2.00
(b) 8.00
1.40
1.60
1.80
2.00 2.20
Time, sec
2.40
2.60
2.80
4.00
3.00
6.00
8.00
1.20
(c)
6.00
2.00 2.20
Time, sec
2.40
2.60
2.80
3.00
Alpha vs time
Analog
Finite difference
4.00
Alpha(rad/s/s), 10
4.00
Alpha(rad/s/s), 10
1.80
6 00
2.00
0
2.00
2.00
0
2.00
4.00
4.00
6.00
6.00
8.00
1.20
1.60
(d ) 8.00
Alpha vs time
Analog
Chebyshev
8.00
1.40
1.40
1.60
1.80
2.00 2.20
Time, sec
2.40
2.60
2.80
3.00
8.00
1.20
1.40
1.60
1.80
2.00 2.20
Time, sec
2.40
2.60
2.80
3.00
Finally, in Figure 3.22d, you can see the accelerometer signal plotted
against angular acceleration as calculated by second-order finite-difference
techniques using raw coordinate data. The plot speaks for itselfthe accelerations are too noisy to mean anything.
3.5
76
KINEMATICS
1
Thigh Angle- q21
q21
2
qKNEE-q21q43
+ve
ve
FOR FLEXION
FOR EXTENSION
Shank Angle-q43
q43
4
5
q65 q
76
qANKLE-q43 q65 + 90
+ve FOR PLANTARFLEXION
ve FOR DORSIFLEXION
qMT-PH-q65 q76
Figure 3.23 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.
y3 y4
x3 x4
(3.13)
yj yi
xj xi
(3.14)
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.
77
ankle angle = a = 43 65 + 90
xi +1 xi 1
m/s
2t
(3.15)
Note that the velocity is at a point halfway between the two samples, as
depicted in Figure 3.24. 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 displacement data in Equation (3.14); the angular acceleration
78
KINEMATICS
Figure 3.24 Finite-difference technique for calculating the slope of a curve at the i th
sample point.
i +1 i 1
rad/s
2t
(3.16)
Vxi +1 Vxi 1
m/s2
2t
(3.17)
Vxi +1/2 =
(3.18a)
Vxi 1/2
(3.18b)
xi +1 2xi + xi 1
m/s2
t 2
(3.18c)
For angular accelerations merely replace displacement data with angular data
in Equations (3.17) or (3.18).
3.6
79
80
KINEMATICS
(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.3).
(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.2, calculate the following
and check your answer from the results tabulated in Table A.3.
(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.
3.7
REFERENCES
3.7 REFERENCES
81
Trenholm, B. G., D. A. Winter, D. Mymin, and E. L. Lansdown, Computer Determination of Left Ventricular Volume Using Videodensitometry, Med. & Biol. Eng. 10:
163173, 1972.
Wells, R. P., D. A. Winter, Assessment of Signal and Noise in the Kinematics of
Normal, Pathological and Sporting Gaits, 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, 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, Computers Biomed. Res. 5: 498504, 1972.
Winter, D. A., H. G. Sidwall, and D. A. Hobson. Measurement and Reduction of Noise
in Kinematics of Locomotion, J. Biomech. 7: 157159, 1974.
Winter, D. A., A. E. Patla, M. G. Ishac, and W. H. Gage. Motor Mechanisms of
Balance During Quiet Standing, J: Electromyogr. Kinesiol. 13: 4956, 2003.
Woltring, H. J. Data Acquisition and Processing Systems in Functional Movement
Analysis. Minerva Orthop. Traumatol. 38: 703716, 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: 318329, 1999.