Vous êtes sur la page 1sur 5

2017 International Conference on Rehabilitation Robotics (ICORR) QEII Centre, London, UK, July 17-20, 2017.

Biomimetic Finger Extension Mechanism for Soft Wearable Hand Rehabilitation Devices

Dong Hyun Kim, Si-Hwan Heo, and Hyung-Soon Park*, Member, IEEE

Abstract For the rehabilitation and assistance of the hand functions, wearable devices have been developed, and the interest in tendon driven mechanisms have especially increased since it allows light weight and compact design. The tendon driven hand rehabilitation devices provides grasping force via exo-tendons routed on the dorsal and palmar sides of the hand pulled by remotely located actuators. However, most of the devices were not able to provide natural joint extension sequence of the finger and showed hyperextension of finger joints because the tendons for extension were fixed at the fingertip, concentrating the torque at the distal interphalangeal joint. In this study, a ring-type biomimetic finger extension mechanism was developed, which mimics the origin, structure, and orientation of the extensor tendon. The biomimetic mechanism was evaluated by comparing the motion with voluntary finger extension and the motion made by other conventional tendon driven finger extension mechanisms. The biomimetic extension mechanism provided the same joint extension sequence with voluntary finger extension, and the fully extended posture was most close to the voluntary finger extension among the tendon- driven mechanisms used in the experiments. The joint angle differences between the proposed tendon mechanism and the voluntary finger extension was -1.2° ± .4°, -2.9°±2.0°, and - 3.1°±8.0° for distal phalangeal, proximal phalangeal, and metacarpo-phalangeal joint, respectively.

KeywordsWearable device; Tendon driven mechanism; Finger extension; Biomimetic design;



For patients with neurological disorders, rehabilitation of the hand function is important to recover the upper-limb function for daily tasks for living. However, the recovery of the hand function is known to be slower than the other joints of the upper-limb[1]. Many hand rehabilitation devices have been developed to provide massive sessions of training of the hand for effective rehabilitation [2-7]. Exoskeleton type devices [2-4] were able to provide complex motion of the hand, but they become bulky and heavy due to the encoders and its mounting structure for hand posture measurement. The bulky design restricts the range of motion (ROM) of the joints and reachability required for conducting various rehabilitation tasks such as moving small objects from one place to another. For compact design around the hand without interrupting the joint ROM, recently, tendon driven mechanisms have drawn attention [5-7]. Among the tendon driven devices, the SEM glove [5] assisted grasping force by pulling exo-tendons

placed along the palmer side of the thumb, middle finger, and ring finger. The Exo-Glove [6] had exo-tendons placed along the dorsal side and palmer side of the fingers, allowing both flexion and extension motion. These devices [5, 6] were able to enhance the strength of the power grasp, but the motions were limited to simultaneous flexion or extension of the distal interphalangeal (DIP), proximal interphalangeal (PIP) and metacarpophalangeal (MCP) joints. Therefore, dexterous pinching grasp tasks were not able to be achieved with these devices [5, 6]. BiomHED [7] introduced two additional exo- tendons mimicking the orientation of the lumbrical and interossei that help to achieve dexterous motion of the finger


While devices using tendon driven mechanisms were successful in assisting grasping force [5-7], and making natural grasping postures [7], it is still challenging to implement natural extension motion of fingers. The origin of the extensor in most of the devices [5-7] were located at the tip (or distal phalange) of the fingers and the tendons were guided through couple of pulleys located on the dorsal aspect of the hand. The angle between force at the tip and the distal phalange increases as the finger extends while the angles between force vector along the pulleys and the middle/proximal phalanges are small and constant. For this reason, most of the tendon-driven mechanisms experience hyperextension at the DIP joint and induce boutonniere deformity at the early phase of finger extension. Furthermore, when the device is used for stroke patients having uncontrolled spasticity and/or contracture [9] and clenched-hand posture [10], the hyperextension of the DIP joint would be more eminent, and the MCP joint would also hyperextend before the PIP joint due to the extra high stiffness at PIP joint. Hyperextension may cause injuries on the joints, and repetitive hyperextension would increase the possibility of osteoarthritis (OA) by providing concentrated load [11, 12]. In addition, the unnatural joint torques provided by the external devices would keep the patients from practicing proper muscle synergies. The patients would try to compensate the unnatural torques, and it would lead to learning distorted pattern of muscle synergies. Even though BiomHED [7] was designed for preventing the hyperextension by co-activation of the tendons for extension and flexion, the tendon force will concentrate along the axial direction of the finger phalanges causing discomfort due to extra high compressional force on the finger joints. In this study, a ring type tendon driven mechanism was designed by mimicking the anatomic structure of the web-

* This research was supported by the convergence technology development program for bionic arm through the National Research Foundation of Korea (NRF) funded by the Ministry of Science, ICT &Future Planning (2015M3C1B2052817). D. H. Kim is with Mechanical Engineering Department of Korea Advanced Institute of Science and Technology, Daejeon, S. Korea 34141. (e- mail: bomdon@kaist.ac.kr).

S. H. Heo is with Mechanical Engineering Department of Korea

Advanced Institute of Science and Technology, Daejeon, S. Korea 34141. (e-

mail: hershey@kaist.ac.kr).

H. S. Park is with Mechanical Engineering Department of Korea

Advanced Institute of Science and Technology, Daejeon, S. Korea 34141. (e-

mail: hyungspark@kaist.ac.kr).

978-1-5386-2296-4/17/$31.00 ©2017 IEEE


shaped finger extensor and using multiple origins (fixed points) to provide natural extension of the finger. The effect of the new design was experimentally evaluated by comparing the finger extension motion created by the proposed mechanism with the natural finger extension obtained from the voluntary finger extension. The performance was also compared with the conventional finger extension mechanisms.


A. Anatomy of the finger extensor

The extensor of the finger has a web-shaped structure and multiple origins to adjust the torque of each joint to make natural extension motion with a certain sequence and the amount of joint angle displacement. First, the extensor has two origins, one originating from the distal phalange and the other from the middle phalange. The tendon originating from the distal phalange splits into two and passes the lateral side of the PIP joint, which is called lateral slip (Figure 1). The two lateral slips eventually meet with the other tendon originating from the middle phalange (central slip) around the dorsal aspect of the proximal phalange. Due to this structure, the torque exerted on the DIP joint varies depending on the PIP joint angle. The lateral slip widens and slips downward (palmer direction) when the PIP joint is flexed, reducing the joint torque at the DIP joint, while it narrows and slips upward (dorsal direction) when the PIP joint is extended, increasing the DIP joint toque. Also, as the extensor originates from two points, the extensor force is distributed to the tip and middle phalange, preventing torque concentration on a single joint. Due to the multiple origin of the extensor and slide-able structure of the lateral slips, the joint extension of the DIP joint and PIP joint follows a certain sequence starting with PIP joint extension, and the DIP joint extension following after. Another special structure of the web-shaped extensor is that the lateral slips are connected with the lumbrical and interrosei tendons (lateral band) around the PIP joint. The lateral bands act as a ligament with some stiffness, balancing the orientation of the lateral slips, and guiding them to slip up and down symmetrically along the side of the PIP joint. Also, when the lumbrical and interrosei is enforced, the DIP joint will extend while the MCP joint flexes. The PIP joint movement depends on the orientation of the lateral slip, which would extend when the lateral slip is placed downward (palmer direction) while it would flex when the lateral slip is placed upward (dorsal direction). In this study, we mimicked the origins and structure of the extensor with exo-tendons placed along the hand surface, which divides into three slips and join near the proximal phalange. To balance the lateral slips, lateral bands were also replicated, but they only functioned as a ligament and they were not actuated.

only functioned as a ligament and they were not actuated. Figure 1. Anatomic structure of the

Figure 1.

Anatomic structure of the finger extensor (upper figure: dorsal view, lower figure: side view)

B. Mechanism Design

To guide the exo-tendons at the biomimetic orientation, ring-type tendon guides were custom designed and prototyped using a 3D printer (Cubicon Style, Hyvision systems Inc., Seongnam, South Korea). The ring-type tendons guides were placed on the middle phalange and proximal phalange. The palmer aspect of the plastic guide was open for easy don/doff of the device. Velcro straps were attached to plastic guides, to fix the plastic guides on the finger. Dyneema (Royal DSM, Heerlen, Netherlands) wires were used for the exo-tendons of the device. Five Dyneema wires were connected with each other by making fixed and slip knots as shown in Figure 2. The origin was fixed to the middle phalange ring and rubber thimble. The central slip exo-tendon (tendon 3, Figure 2.) was guided along the dorsal aspect of the MCP joint, while the lateral band exo-tendons (tendon 1and 2) were guided along the palmer aspect of the MCP joint to mimic the structure and orientation of the extensor and lumbrical shown in Figure 1. Around the MCP joint, the exo-tendons were guided with teflon tubes inside a custom designed silicon glove structure made with KE-1300T silicon (Shin-Etsu Chemical Co., Ltd., Tokyo, Japan).

silicon (Shin-Etsu Chemical Co., Ltd., Tokyo, Japan). Figure 2. Schematic of extensor exo-tendon structure of the

Figure 2.

Schematic of extensor exo-tendon structure of the biomimetic finger extension mechanism (Dorsal view)

In this study, the exo-tendon that replicates the extensor (exo-tendon 3 in Figure 2) was enforced manually for the experimental evaluation, and exo-tendons that replicate the lumbrical (exo-tendon 1 and 2 in Figure 2) were restricted at a certain length by fixing the end to the silicon glove without actuation.



A. Experimental Protocol

To evaluate how natural extension motion could be made by the biomimetic extension mechanism, the motion made by pulling the exo-tendon for extension was compared with voluntary finger extension and the motion made by two other extension mechanisms with different tendon routings. The other mechanisms consisted of a single tendon that pass along the dorsal aspect of the finger. The fixed point of extension tendon for one mechanism was on the tip (Mech-I) and for the other mechanism, it was on the middle phalange (Mech-II). The overview of the compared devices are shown in Figure 3. Experimental validation was conducted for a single male subject (28 years old) who gave written consent approved by IRB (KH2015-10) at Korea Advanced Institute of Science and Technology.

at Korea Advanced Institute of Science and Technology. Figure 3. (a) (b) (c) Overview of the

Figure 3.




Overview of the compared mechanisms: (a) Mech-I, extensor

fixed at the tip, (b) Mech-II, extensor fixed at the middle phalange, (c)

Biomimetic extension mechanism.

The motion of the finger was captured by using a Vicon motion capture system consisting of two T160 camera and five T40s cameras (Vicon Motion Systems Ltd., Oxford, UK). The 3-dimensional marker position data was collected at a 100 Hz sampling rate. The markers were attached to the hand to record the movement of the finger. The marker setup is shown in Figure 4. A total of seven makers were placed on the dorsal aspect of the tip (P tip ), DIP joint (P dip ), PIP joint (P pip ), MCP joint of the index finger (P mcp1 ), MCP joint of the middle finger (P mcp2 ), and one along the index metacarpal bone (P meta ).

and one along the index metacarpal bone (P m e t a ). Figure 4. Marker

Figure 4.

Marker positions and definition of joint flexion angle

After the data collection, the joint angle trajectories of DIP, PIP, and MCP joint were computed based on the 3D marker position with a custom Matlab program (Mathworks Inc., Natick, Massachusetts). The hand plane (S hand ) was defined by three points P mcp1 , P mcp2 , and P meta as shown in Figure 4. The x-

axis was defined as the unit vector of P ⃗⃗⃗⃗⃗⃗⃗⃗⃗⃗⃗⃗⃗⃗⃗⃗⃗⃗⃗⃗⃗⃗ meta P mcp1 , the y-axis was defined as the normal vector of S hand toward the palmer direction, and the z-axis was defined as the vector normal to the x and y axis. The sagittal plane (S z ) was defined as the plane normal to the z-axis. To define the finger joint angles, ⃗⃗⃗⃗⃗⃗⃗⃗⃗⃗⃗⃗⃗⃗⃗⃗

vectors, corresponding to

each phalange of the finger, were projected to the S z plane ⃗⃗⃗⃗⃗⃗⃗⃗⃗⃗⃗⃗⃗⃗⃗⃗

′) . The angle between


was defined as the MCP angle, the

was defined as the



was defined as the DIP flexion angle. The vectors

projected to the sagittal plane(S z ) were utilized to compute the

joint angles, because when the markers were not attached in a straight line along the dorsal aspect of the finger, the angles ⃗⃗⃗⃗⃗⃗⃗⃗⃗⃗⃗⃗⃗⃗⃗⃗


directly had an offset angle even when the finger was extended straight to 0°. In the voluntary finger extension experiment, the subject was asked to initially set the hand at a neutral posture with the muscles relaxed, and when the motion capture started, the subject extended the index finger. During the experiments conducted with extension mechanisms, the subject was also asked to initially set the hand at a neutral posture with the muscles relaxed, and when the motion capture started, the exo-tendon for extension was pulled manually from the experimenter by gradually increasing force with a slow speed to neglect the inertial effect. The experimenter pulled the exo- tendon until the finger was fully extended, and when excessive hyperextension of the finger joints were present the experimenter stopped pulling the extensor exo-tendon. For each type of task (voluntary extension, extension by Mech-I, Mech-II, and the biomimetic extension mechanism), the experiment was repeated ten times.


PIP flexion angle, and the angle between ⃗⃗⃗⃗⃗⃗⃗⃗⃗⃗⃗⃗⃗⃗⃗⃗ P PIP P DIP



theP mcp1 P PIP , P PIP P DIP


( P mcp1 P PIP , P PIP P DIP

meta P mcp1








and P mcp1 P PIP

angle between P ⃗⃗⃗⃗⃗⃗⃗⃗⃗⃗⃗⃗⃗⃗⃗⃗⃗⃗⃗ mcp1 P PIP




P mcp1 P PIP










The joint angle trajectories of the DIP, PIP, and MCP joint were computed for all the trials, and they were filtered with the 3 rd order butterworth filter function in Matlab at the cutoff frequency of 1 Hz, because the finger movement of interest was slow. The first and last time frame of interest was decided based on the joint velocity. For each joint, the time frame when the joint velocity was over a certain threshold (10 degrees/sec for voluntary extension experiment and 3 degrees/sec for device experiments) was investigated, and the time region between the first and last time frame was set as the region of interest. The time region of interest was normalized, and for each time step, the average and standard deviation of all trials with the same type of task were derived. The average joint angle trajectories of each joint for voluntary extension, and its comparison with motions made by Mech-I, Mech-II, and biomimetic extension mechanism are shown in Figure 5.


(a) (b) Figure 5. (c) (d) Joint angle trajectories, and start and end timing of



(a) (b) Figure 5. (c) (d) Joint angle trajectories, and start and end timing of extension:

Figure 5.



Joint angle trajectories, and start and end timing of extension:

(a) voluntary extension; (b) Mech-I; (c) Mech-II; (d) Biomimetic extension mechanism. In (b)-(d), the blue vertical lines indicate the timing of

voluntary motion.

The colored filled line indicates the average of each moment and the shaded area indicates the standard deviation from the average. The black vertical line in the figure indicates the moment when the joint has moved 5° and the moment when 5° are left until full extension. These timings were set as the starting and ending time of the joint extension. From the result of the voluntary motion, we could notice that the starting sequence of the joint movement is PIP joint, DIP joint, MCP joint, and the ending sequence is DIP joint, PIP joint, MCP joint. The joint angle when the finger is fully extended is 0.7 ° ± 0.5°, -0.9° ± 0.6°, and 9.9° ± 3.0° for the DIP, PIP and MCP joint. The starting and ending sequence of the joint trajectories made by the Mech-I were the same with the voluntary extension motion. However, for the Mech-I, the DIP joint started to hyperextend (under 0°) when the PIP joint extended only 50% of the full range of extension made by the mechanism, and the fully extended angle of the DIP joint was -13° ± 4.0. Also, large standard deviations were shown for the MCP joint in the second half of the extension phase. This was due to the regulation of extension force and ending time to prevent excessive torque on the DIP and PIP joint. In the case of Mech-II, the starting sequence was the same as the voluntary extension while the ending sequence was different, with a sequence of MCP joint, DIP joint, PIP joint. There was no hyperextension made by the Mech-II, but the final angle of the DIP joint was 10° ± 1.4°, which is larger than that of voluntary motion. The biomimetic extension mechanism showed the same starting and ending sequence of the joint movement with the voluntary extension motion, and almost no hyperextension occurred (under -5°). Also the final joint angle of the DIP joint was -1.7° ±2.4°, which was the closest value with that of the voluntary extension among the three

mechanisms. The difference of the joint angle between the voluntary motion and the extension motion made by the mechanisms are arranged in Table I. The difference was calculated by subtracting the fully extended joint angle made by the mechanisms from that of the voluntary extension.



Mechanism Joint angle difference with voluntary extension( ° ) PIP joint DIP joint MCP joint


Joint angle difference with voluntary extension(°)

PIP joint

DIP joint

MCP joint

extension( ° ) PIP joint DIP joint MCP joint Mech-I -14 ° ± 4.0° - 2.4
extension( ° ) PIP joint DIP joint MCP joint Mech-I -14 ° ± 4.0° - 2.4


-14° ± 4.0°-

2.4° ± 6.6°

7.3° ± 11°

Mech-I -14 ° ± 4.0° - 2.4 ° ± 6.6° 7.3 ° ± 11° Mech-II Biomimetic




9.3° ± 1.5°

-1.2° ± 3.4°

4.3° ± 4.7°

5.8° ± 4.7°



4.3 ° ± 4.7° 5.8 ° ± 4.7° -2.9°±2.0° -3.1°±8.0°
Mechanism 9.3 ° ± 1.5° -1.2 ° ± 3 .4 ° 4.3 ° ± 4.7° 5.8


In this study, a ring-type biomimetic extension mechanism was developed by replicating the structure, origin and orientation of the web-like extensor. To evaluate the mechanism, the joint angle trajectory made by the biomimetic extension mechanism was compared with the joint angle trajectories from voluntary extension and other conventional mechanisms. The biomimetic mechanism and Mech-I showed the same starting and ending joint extension sequence with voluntary extension, while for Mech-II, the ending time of the MCP joint movement was the earliest. Mech-I hyperextended the DIP joint, while Mech-II did not fully extend the DIP joint as much as the voluntary extension. The fully extended DIP joint angle made by the biomimetic extension mechanism, however, was similar with that of the voluntary extension. Also, the final extended posture made by the biomimetic mechanism was the closest with the voluntary motion (Table I). From this result, we found that the biomimetic mechanism could make more natural motions than conventional mechanisms with a single fixed point. In addition, we could notice that the number and position of the tendon fixed point and the structure of the tendon were important in generating natural sequence and amount of extension of the finger joints. Preventing hyperextension and making natural joint extension sequence could also be achieved by including a stopper mechanism on the dorsal aspect [13], which restricts joint extension, or by including springs to adjust the stiffness of the joint. However, the stopper and spring mechanisms around the hand would make the device bulky, restricting the ROM of the joints. The biomimetic finger extension mechanism, however, prevented extensive hyperextension and provided natural joint extension sequence, while retaining the compact design. The proposed design achieved a natural sequence of the joint extension, but the ending time of the MCP joint extension was early compared to the voluntary extension (Figure 5). The relatively later ending of the extension of the PIP joint compared to the natural sequence in voluntary motion is also caused by the early ending time of the MCP joint extension. Mech-I and Mech-II also showed early timings. This phenomenon is due to the lack of MCP joint extension regulation from lateral band (lumbrical and interrosei) enforcement existing in voluntary movements. Intrinsic muscles are involved to provide tension at the lateral band in the finger, but the proposed design only used the lateral band as a ligament without actuation. It is expected that, if the lateral bands in the proposed design are actuated, the MCP joint movement could be regulated using the same extensor


structure. The adjustability of the complex extensor structure, however, should be improved to adjust the force distribution for patients with neurological disorders, having different joint stiffness. In this study, the preliminary results for a single subject have been investigated. For generalization of the results, a follow up study should be conducted with an extended number of subjects. Also, to evaluate the efficacy of the device when it is used for patients with neurological disease such as stroke and spinal cord injury, the device must also be tested for the patients. The ring-type biomimetic extension mechanism is expected to provide natural extension of the finger extensor for the hand rehabilitation of assistance devices, preventing injuries from hyperextension of the joint. Also by providing the proper extension joint torque coordination, it is expected to be helpful for correct rehabilitation of the muscle synergies in conducting various grasping tasks. Furthermore, if the lumbrical exo-tendons in the biomimetic mechanism are also regulated passively by adjusting the stiffness or actively by controlling tension by a servo-motors, it is expected to provide natural joint extension sequence to patients who have contracture, usually severe on the PIP joints [10].


[1] T. E. Twitchell, "The restoration of motor function following


hemiplegia in man," Brain, vol. 74, no. 4, pp. 443-80, Dec 1951. S. Ueki et al., "Development of a Hand-Assist Robot With Multi- Degrees-of-Freedom for Rehabilitation Therapy," (in English), Ieee-

Asme Transactions on Mechatronics, vol. 17, no. 1, pp. 136-146, Feb



C. N. Schabowsky, S. B. Godfrey, R. J. Holley, and P. S. Lum,



"Development and pilot testing of HEXORR: Hand EXOskeleton Rehabilitation Robot," Journal of NeuroEngineering and Rehabilitation, journal article vol. 7, no. 1, p. 36, 2010. A. Wege and G. Hommel, "Development and control of a hand

exoskeleton for rehabilitation of hand injuries," in 2005 IEEE/RSJ International Conference on Intelligent Robots and Systems, 2005, pp.


M. Nilsson, J. Ingvast, J. Wikander, and H. v. Holst, "The Soft Extra

hand," Intelligent Service Robotics, journal article vol. 6, no. 4, pp.



Muscle system for improving the grasping capability in neurological rehabilitation," in 2012 IEEE-EMBS Conference on Biomedical Engineering and Sciences, 2012, pp. 412-417. U. Jeong, H.-K. In, and K.-J. Cho, "Implementation of various control


algorithms for hand rehabilitation exercise using wearable robotic

181-189, 2013. S. W. Lee, K. A. Landers, and H. S. Park, "Development of a


biomimetic hand exotendon device (BiomHED) for restoration of functional hand movement post-stroke," IEEE Trans Neural Syst Rehabil Eng, vol. 22, no. 4, pp. 886-98, Jul 2014. K. Kursa, L. Lattanza, E. Diao, and D. Rempel, "In vivo flexor tendon

forces increase with finger and wrist flexion during active finger flexion and extension," J Orthop Res, vol. 24, no. 4, pp. 763-9, Apr


D. G. Kamper, H. C. Fischer, E. G. Cruz, and W. Z. Rymer, "Weakness is the primary contributor to finger impairment in chronic stroke," Arch Phys Med Rehabil, vol. 87, no. 9, pp. 1262-9, Sep 2006.


C. A. Trombly, Occupational Therapy for Physical Dysfunction, 3 ed. Baltimor, MD: Lippincott Williams & Wilkins, 1989.


D. T. Felson, "Osteoarthritis as a disease of mechanics," Osteoarthritis and cartilage, vol. 21, no. 1, pp. 10-15, 2013.

[12] T. A. Maxian, T. D. Brown, and S. L. Weinstein, "Chronic stress tolerance levels for human articular cartilage: two nonuniform contact models applied to long-term follow-up of CDH," Journal of biomechanics, vol. 28, no. 2, pp. 159-166, 1995.


K. M. Triandafilou, J. Ochoa, X. Kang, H. C. Fischer, M. E. Stoykov, and D. G. Kamper, "Transient impact of prolonged versus repetitive stretch on hand motor control in chronic stroke," Topics in stroke rehabilitation, 2015.