2011 IEEE International Conference on Rehabilitation Robotics

Rehab Week Zurich, ETH Zurich Science City, Switzerland, June 29 - July 1, 2011

An EMG-driven Exoskeleton Hand Robotic Training

Device on Chronic Stroke Subjects
Task Training System for Stroke Rehabilitation

N.S.K. Ho, K.Y. Tong, Senior Member, IEEE, X.L. Hu, K.L. Fung, X.J. Wei, W. Rong, E.A. Susanto
Department of Health Technology and Informatics
The Hong Kong Polytechnic University
Hong Kong SAR, China
Email: [email protected]

Abstract—An exoskeleton hand robotic training device is 15]. Only few robotic devices for the hand rehabilitation are
specially designed for persons after stroke to provide training on found in the market [16-20]. Therefore we have started this
their impaired hand by using an exoskeleton robotic hand which project to develop an interactive rehabilitation robot for hand
is actively driven by their own muscle signals. It detects the functions task training to meet the needs of the stroke
stroke person’s intention using his/her surface electromyography rehabilitation.
(EMG) signals from the hemiplegic side and assists in hand
opening or hand closing functional tasks. The robotic system is
made up of an embedded controller and a robotic hand module II. EXOSKELETON HAND ROBOTIC TRAINING DEVICE
which can be adjusted to fit for different finger length. Eight
chronic stroke subjects had been recruited to evaluate the effects We have specially designed an exoskeleton hand robotic
of this device. The preliminary results showed significant training device for person after stroke to actively train their
improvement in hand functions (ARAT) and upper limb impaired hand functions. By measuring his/her surface
functions (FMA) after 20 sessions of robot-assisted hand electromyography (EMG) signals from the impaired hand
functions task training. With the use of this light and portable muscles, this robotic device detects the stroke person’s
robotic device, stroke patients can now practice more easily for intention and assists in hand opening or hand closing.
the opening and closing of their hands at their own will, and
handle functional daily living tasks at ease. A video is included
together with this paper to give a demonstration of the hand
robotic system on chronic stroke subjects and it will be presented
in the conference.

Keywords- exoskeleton; rehabilitation; robot; hand; stroke;

I. INTRODUCTION
Stroke is one of the diseases which leads to high disability
and death according to the World Health Organization [1].
Often the stroke subjects’ motor functions and mobility are
greatly affected [2, 3]. Approximate 70 to 80 percent of the
A
stroke survivors require long term medical care [4, 5] and live
with a poor quality of life (QOL) [6, 7].
Some of the stroke survivors who completed a
rehabilitation program for the upper extremities were able to
recover some of the proximal motor functions at the shoulder
and elbow joints but limited recovery for the hand and wrist
joints [8, 9]. Hand functions such as hand opening and closing
are useful for many daily tasks but it has been a challenge to
develop an effective training device for the hand functions
rehabilitation. B C
Robots have proved to be effective in assisting the therapist
to provide safe and intensive rehabilitation training for the Figure 1. Exoskeleton hand robotic training device: A) robotic hand module
stroke subjects [10, 11]. Rehabilitation robot for elbow and with 5 linear actuators at the back, B) hand is secured with the robotic hand
wrist has already been proven effective in clinical trials [11- module using Velcro straps and C) holding an object with robotic hand

This work was supported by the General Research Fund (GRF) of the
Research Grants Council of Hong Kong SAR, China (PolyU 5292/08E) and
Innovation and Technology Fund (GHP/003/07)

978-1-4244-9862-8/11/$26.00 ©2011 IEEE

 The system consists of a robotic hand module (see figure 1) An embedded controller is built to control the robotic hand
and an embedded controller. The fingers and the palm of the module and monitor the surface EMG signals for hand opening
stroke subject hand are mounted comfortably on to the robotic and closing tasks. Microchip microcontroller is used to control
hand module using finger rings and Velcro straps. The palm all the linear actuators and measure the surface EMG signals
area of the hand and DIP joint on figure are left open to allow from the two major muscle groups – abductor pollicis brevis
user to grip and feel the objects with their own fingers. (APB) and extensor digitorum (ED). EMG signal from the
APB is used to control the hand closing task while the signal
from the ED is used to control the hand opening task. Both
A. Robotic hand module channels of EMG signals (APB & ED) are sampled at 1 kHz.
The embedded controller contains a wireless module for
wireless communication. Therefore, the therapist can configure
Each hand module consists of five finger assemblies and a
the exoskeleton robotic hand and choose different training
palm support platform. Each finger assembly is actuated by a
modes using the wireless remote control system (see figure 3).
single linear actuator (Firgelli L12) and provides 2 degrees of
freedom (DOFs) for each finger at the MCP and PIP together
by the mechanical linkage system. The proximal section rotates
around the virtual centre located at the proximal
interphalangeal (PIP) joint whereas the distal section rotates
around the virtual centre located at the metacarpophalangeal
(MCP) joint (see figure 2). From fully extended position to
fully flexed position, the finger assembly provides 55 degrees
and 65 degrees range of motion (ROM) for the MCP and PIP
joints respectively. When under no load, the maximum
contraction speed of the robotic hand is approximately
2seconds to fully open or close the robotic hand.

Figure 3. Full set of exoskeleton hand robotic training device

The whole system is designed to be portable and self

contained allowing the stroke patient to carry it around to
practice daily activities.

III. METHODOLOGY

Eight chronic stroke subjects were recruited to evaluate the

exoskeleton hand robotic training device with hand function
task training. The evaluation was conducted at the department
of Health Technology and Informatics, the Hong Kong
Polytechnic University.
Figure 2. Movement of finger assembly around the virtual centres A. Training set up
All finger assemblies are identical and each can be adjusted
to fit for different finger length and align the virtual centre of Stroke subjects would sit comfortably in front of a table to
rotation to the MCP and PIP joints, thus provides a natural do the task training. When fitting the robotic hand module with
movement of finger flexion and extension. Including all 5 the impaired hand of the stroke subjects, Velcro straps were
finger assemblies, 5 linear actuators and the palm support used to securely hold the hand in place and mount individual
platform, the total weight of the robotic hand module is only fingers on the finger assemblies. On average, it took 30 seconds
500 grams. to don the exoskeleton. Surface EMG electrodes (Ambu Blue
Sensor SE-00-S/50) were used to collect the surface EMG
B. Embedded controller signals from the extensor digitorum (ED) and abductor pollicis
brevis (APB) (see figure 4).
 C. Hand function task training

In this evaluation each subject was required to attend 20

training sessions with training intensity about 3-5 sessions per
week. Clinical outcome measures were assessed before and
after the 20 sessions.

During the training session, the stroke subjects were

required to complete 2 sets of upper limb training tasks assisted
by the exoskeleton hand robotic training device, each with 10
minutes training period. First task required stroke subject to
move an object, which is a foam, across a table horizontally for
a 50cm distance. And the second task required stroke subject to
move an the foam vertically above the table for 20cm. In each
10 minutes, the stroke subjects repeated the same task with
hand opening, grasp the foam, move to the target position and
then release the foam at their comfortable speed (see figure 6).
Figure 4. Donning process for the wearing the robotic hand
And there was a 5 minutes rest between these two tasks.

B. EMG Control strategy

The control strategy used in the training was the EMG-

triggered training mode. A threshold of 20% of the maximum
voluntary contraction (MVC) EMG signals was used to trigger
the hand opening and hand closing motions. Therefore,
baseline and maximum voluntary contraction (MVC) of the
EMG signals were measured at the beginning of each training
session. When the robotic system was running in a hand
closing triggering mode, it would wait until the EMG signals
from the APB muscle exceeded the 20% of its MVC value
before starting the hand closing action. And when it was
running in a hand opening triggering mode, it would wait until
the EMG signals from the ED muscle exceeded the 20% of its
MVC value before starting the hand opening action. See figure
5.

Figure 6. Vertical upper limb training task

D. Stroke subjects

Eight chronic stroke subjects were recruited to evaluate the

effectiveness of the hand functions task training using the
exoskeleton hand robotic training device. All attended 20 hand
Figure 5. EMG signals with EMG-triggered status
functions task training sessions. The demographic data of the rehabilitation and reported the clinical results after 20 session
recruited subjects is shown in Table 1. of training on 8 chronic stroke subjects. The robotic device
encourages the stroke subjects to use their impaired hand
TABLE I. DEMOGRAPHIC INFORMATION
muscles to control the robotic hand module. With the
repeatedly open and close functional task training, the
preliminary data showed the motor functions on the hand and
Subject Affected Limb Time after
Gender Age Stroke Type
Side Stroke
upper limb had been improved after 20-session training. A
larger clinical was needed in future clinical studies.
Chronic
1 F 28 Hemo. Left 4 years REFERENCES
2 M 42 Hemo. Right 14 years
3 F 63 Isch. Left 4 years [1] J. Mackay and G. Mensah, The Atlas of Heart Disease and Stroke.
4 M 67 Hemo. Right 14 years Geneva, Switzerland:World Health Organization, 2004 [Online].
5 M 49 Isch. Left 8 years Available:
http://www.who.int/cardiovascular_diseases/resources/atlas/en/
6 M 58 Isch. Left 3 years
[2] S. Rathore, A. Hinn, L. Cooper, H. Tyroler, and W. Rosamond,
7 M 64 Isch. Right 4 years "Characterization of incident stroke signs and sympotoms: findings from
8 M 51 Isch. Left 10 years the atherosclerosis risk in communities study," Stroke, vol. 33, pp. 2718-
Female (F); Male (M); Hemorrhage (Hemo); and Ischemia (Isch). 2721, 2002.
[3] G. Gresham, P. Duncan, W. Stason, H. Adams, A. Adelman, and D.
Alexander, "Post-stroke rehabilitation," Department of Health and
Human services. Public Health Service, Agency for Health Care Policy
E. Outcome measurements and Research, Rockville, M.D. 1995.
[4] V. M. Parker, D. T. Wade, and R. L. Hewer, "Loss of arm function after
Clinical assessments such as Fugl-Meyer Assessment stroke: measurement, frequency, and recovery," Int Rehabil Med, vol. 8,
pp. 69-73, 1986.
(FMA) and Action Research Arm Test (ARAT) were used to
[5] H. Nakayama, H. S. Jorgensen, H. O. Raaschou, and T. S. Olsen,
evaluate the motor improvement after the robot-assisted hand "Recovery of upper extremity function in stroke patients: The
functions task training. FMA which evaluates the voluntary Copenhagen stroke study," Arch Phys Med Rehabil, vol. 75, pp. 394-
motor functions of the upper limb contains two separate 398, 1994.
measurements of the shoulder and elbow section (S&E) and the [6] T. B.Wyller, J. Holmen, P. Laake, and K. Laake, “Correlates of
wrist and hand section (W&H). ARAT focuses on the subjective well-being in stroke patients,” Stroke, vol. 29, pp. 363–367,
measurement of the hand functions based on common daily 1998.
activity tasks such as picking small objects like marble for [7] E. J. Jonkman, A. W. de Weerd, and N. L. H. Vrijens, “Quality of life
pinching or holding a tennis ball for grasping. after first ischemic stroke. longterm developments and correlations with
changes in neurological deficits, mood and cognitive impairment,” Acta.
Neurol. Scand., vol. 98, pp. 169–175, 1998.
IV. RESULTS [8] J. Chae, G. Yang, B. K. Park, and I. Labatia, "Muscle weakness and
cocontraction in upper limb hemiparesis: relationship to motor
impairment and physical disability," Neurorehabilitation & Neural
Table 2 provides a statistical summary of the training Repair, vol. 16, pp. 241-248, 2002.
effects on the subjects before and after the 20-session hand [9] J. Chae and R. Hart, "Intramuscular hand neuroprosthesis for chronic
robot training. It was found that there are significant motor stroke survivors," Neurorehabilitation & Neural Repair, vol. 17, pp. 109-
improvements in the FMA-S&E, and FMA-W&H, and ARAT 117, 2003.
clinical assessments. The improvement in ARAT score reflects [10] R. Colombo, F. Pisano, S. Micera, A. Mazzone, C. Delconte, M. C.
Carrozza, P. Dario, and G. Minuco, "Robotic techniques for upper limb
the motor recovery in hand and finger functions. The increased evaluation and rehabilitation of stroke patients," IEEE T Neur Sys Reh,
FMA scores suggest there is motor improvement in the whole vol. 13, pp. 311-323, 2005.
upper limb after the training. [11] K. Y. Tong and X. L. Hu, "Service Robotics: Robot-Assisted Training
for Stroke Rehabilitation," in Service Robotics. Vienna, Austria: I-Tech
Education and Publishing, 2008.
TABLE II. STATISTICAL SUMMARY
[12] R. Song, K. Y. Tong, X. L. Hu, and X. J. Zheng, "Myoelectrically
Assessments Mean ± SD Paired t-test controlled robotic system that provide voluntary mechanical help for
(max. score) Pre Post persons after stroke," presented at IEEE 10th International Conference
FMA-S&E (42) 21.50±2.59 28.00±7.40 0.0426* on Rehabilitation robotics, Noordwijk, the Netherlands, 2007.
FMA-W&H (24) 7.83±5.11 11.83±5.85 0.0035* [13] R. Song, K. Y. Tong, X. L. Hu, and L. Li, "Assistive control system
ARAT (57) 21.833±8.66 29.67±9.14 0.0121* using continuous myoelectric signal in robot-aided arm training for
* Mean value changes with statistical significance (P<0.05, paired t-test).
patients after stroke," IEEE Transactions on Neural Systems and
Rehabilitation Engineering, vol. 16, pp. 371-379, 2008.
[14] X. Hu, K. Y. Tong, R. Song, V. S. Tsang, P. O. Leung, and L. Li,
"Variation of muscle coactivation patterns in chronic stroke during
V. CONCLUSION robot-assisted elbow training," Arch Phys Med Rehabil, vol. 88, pp.
1022-1029, 2007.
[15] X. L. Hu, K. Y. Tong, R. Song, X. J. Zheng, K. H. Lui, W. W. F. Leung,
This study showed a newly design EMG-driven S. Ng, and S. S. Y. Au-Yeung, "Quantitative evaluation of motor
exoskeleton hand robotic training device for stroke
 functional recovery process in chronic stroke patients during robot- [19] O. Lambercy, L. Dovat, H. Yun, S. Wee, C. Kuah, K. Chua, R. Gassert,
assisted wrist training," J Electromyography Kinesiol, vol. 18, 2008. T. Milner, C. Teo, and E. Burdet, “Rehabilitation of Grasping and
[16] A. Wege, G. Hommel, “Development and Control of a Hand Forearm Pronation/Supination with the Haptic Knob,” in IEEE
Exoskeleton for Rehabilitation of Hand Injuries.” Proc. of the 2005 International Conference on Rehabilitation Robotics (ICORR 2009), pp.
IEEE/RSJ Int. Conf. On Intelligent Robots and Systems (IROS), pp. 22–27.
3461-3466, [20] L. Dovat, O. Lambercy, R. Gassert, T. Maeder, T. Milner, C. Teo and E.
[17] The Amadeo® System, Tyromotion. [Online]. Available: Burdet, “HandCARE: A cable-actuated rehabilitation system to train
http://www.tyromotion.com/en/products/amadeo/ hand function after stroke,” IEEE Transaction in Neural Systems and
Rehabilitation Engineering (2008) 16(6), pp. 582–591.
[18] H. Kawasaki, S. Ito, Y. Ishigure, Y. Nishimoto, T. Aoki, T. Mouri, H.
Sakaeda, and M. Abe, “Development of a hand motion assist robot for
rehabilitation therapy by patient self-motion control,” in Proc. IEEE Int.
Conf. Robotic Rehabil. (ICORR), Jun. 2007, pp. 234–240.