INTERNATIONAL JOURNAL OF PRECISION ENGINEERING AND MANUFACTURING Vol. 13, No. 5, pp.

807-824 MAY 2012 / 807

DOI: 10.1007/s12541-012-0107-2

Current Hand Exoskeleton Technologies for

Rehabilitation and Assistive Engineering
Pilwon Heo1, Gwang Min Gu1, Soo-jin Lee2, Kyehan Rhee2 and Jung Kim1,#
1 Department of Mechanical Engineering, Korea Advanced Institute of Science and Technology, 291 Daehak-ro, Yuseong-gu, Daejeon, Republic of Korea, 305-701
2 Department of Mechanical Engineering, College of Engineering, Myongji University, 116 Myongji-ro, Choin-gu, Yongin, Gyeonggi-do, Republic of Korea, 449-728
# Corresponding Author / E-mail: [email protected], TEL: +82-42-350-3231, FAX: +82-42-350-5230

KEYWORDS: Hand exoskeleton, Rehabilitation, Assistance

In this paper, we present a comprehensive review of hand exoskeleton technologies for rehabilitation and assistive
engineering, from basic hand biomechanics to actuator technologies. Because of rapid advances in mechanical designs and
control algorithms for electro-mechanical systems, exoskeleton devices have been developed significantly, but are still
limited to use in larger body areas such as upper and lower limbs. However, because of their requirements for smaller size
and rich tactile sensing capabilities, hand exoskeletons still face many challenges in many technical areas, including hand
biomechanics, neurophysiology, rehabilitation, actuators and sensors, physical human-robot interactions and ergonomics.
This paper reviews the state-of-the-art of active hand exoskeletons for applications in the areas of rehabilitation and
assistive robotics. The main requirements of these hand exoskeleton devices are also identified and the mechanical designs
of existing devices are classified. The challenges facing an active hand exoskeleton robot are also discussed.

Manuscript received: March 7, 2012 / Accepted: April 15, 2012

1. Introduction of the patient’s performance and progress is difficult with manual

therapy. The efforts to overcome the inefficiency of conventional
Because of their inherent motor and sensory requirements, hand therapy have been realized by robotic rehabilitation. It has been
exoskeleton technologies for rehabilitation and assistive shown that robotic repetitive movement training might be a more
engineering have not progressed as rapidly as the exoskeleton effective treatment, especially for patients who have difficulty in
robots and devices for lower and upper limbs that have become performing unassisted repetitive motion.4 These robotic
popular over the last decade. These requirements have inspired rehabilitation systems can provide effective repetitive training for
considerable developments in robotic hands in terms of their rehabilitation without significantly increasing the costs. The robotic
degrees of freedom, weight, size and dexterous manipulation system can also be used to evaluate the progress quantitatively.
capabilities. At the same time, enhancement of hand functions using These advantages make the use of hand exoskeletons for
exoskeleton technologies for those who have lost or weakened hand rehabilitation applications look promising.
capabilities because of neuromuscular diseases or aging has become Even after an intensive rehabilitation process, hand function
an important issue, because hand functionality is a dominant factor may not be recovered fully. In fact, up to 66% of hemiplegic stroke
in living an independent and healthy life. patients have not regained the function of the paretic arm when
From the viewpoint of rehabilitation after a stroke, it is measured 6 months after the stroke, while only 5% to 20% of
important for the patient to take intensive and continuous patients show complete functional recovery.5-8 Hand exoskeletons
therapeutic exercise for successful rehabilitation. It is shown that can be used to assist the patients who have suffered permanently
recovery from a brain injury is greatly influenced by the lost or weakened hand function.
sensorimotor experience after the injury.1 Highly repetitive training Also, people whose work requires the exertion of a forceful and
can also help to recover the motor function.2,3 However, repetitive hand gripping action are exposed to a high likelihood of
conventional therapy for stroke rehabilitation requires manual developing a musculoskeletal disorder. Therefore, to prevent such
interaction with physical therapists that make the procedure work-related musculoskeletal disorders, it is important to reduce the
labor‐intensive and raise the costs. Also, the quantitative evaluation physical burden on these workers. Hand exoskeletons can be used

© KSPE and Springer 2012

 808 / MAY 2012 INTERNATIONAL JOURNAL OF PRECISION ENGINEERING AND MANUFACTURING Vol. 13, No. 5

to assist the hand function by amplifying the hand gripping force or

automating the motion. Applicable areas include heavy industry,
construction, military, and logistics.
In the following section, the biomechanics of the hand are
discussed and the requirements for the exoskeleton devices are
presented. In Section 3, hand exoskeletons for rehabilitation and
assistance applications that have been developed or are under
development are introduced. Actuator technologies and intention
sensing methods are discussed in Sections 4 and 5, respectively.
Finally, Section 6 summarizes the article and briefly discusses the
challenges facing hand exoskeleton development.

2. Hand Biomechanics

2.1 Anatomy of the Hand

Because a mechanism of a hand exoskeleton is closely coupled
with a hand when it is worn, developing the hand exoskeleton
requires an understanding of hand anatomy and biomechanics for
ensuring safe and effective operation. Specifically, considering the
DOF (degree of freedom) and ROM (range of motion) of each joint Fig. 1 Bones and joints of a human hand
is important for the design of mechanically safe structure. In
addition, the hand movement is complexly related to the intrinsic interphalangeal (IP) joint. They are both bicondylar joints with
and the extrinsic muscles as well as the connective tissues. subsequently greater congruency between the bony surfaces, and
Therefore the systematic knowledge helps achieving proper have one degree of freedom. The transverse diameters of the IP
functions for rehabilitation and assistance. joints are greater than their antero-posterior diameters and the thick
collateral ligaments are tight in all positions during flexion, contrary
2.1.1 Bones and Joints to those in the MCP joint.12 Although the IP joints are frequently
The bones of the hand are naturally grouped into the carpus, modeled and assumed as having single axis of rotation for
comprising the eight bones which make up the wrist and root of the simplicity, in fact they do not remain constant during flexion and
hands, and the digits, each of which is composed of its metacarpal extension.13
and phalangeal segments. The five digits are named as follows from The different shapes of the finger joints result in varying DOF
the radial to the ulnar side: thumb, index finger, middle finger, ring at each joint. Also, the orientation of the thumb and the unique
finger, and little finger. Each finger ray is composed of one configuration of its CMC joint provide this digit with a large range
metacarpal and three phalanges, except for the thumb (which has of motion and greater flexibility.14,15 The wrist is extended 20° in
two phalanages). There are 19 bones and 14 joints distal to the neutral radial/ulnar deviation at the resting posture. The resting
carpals, as shown in Fig. 1. The carpal bones are arranged in two posture is a position of equilibrium without active muscle
rows, with those in the more proximal row articulating with the contraction. The MCP joints are flexed approximately 45°, the PIP
radius and ulna. Between the two is the intercarpal articulation. joints are flexed between 30°and 45°, and the DIP joints are flexed
Each finger articulates proximally with a particular carpal bone at between 10° and 20° at the resting posture. Flexion of the MCP
the carpometacarpal (CMC) joint. The CMC joint of the thumb is a joints is approximately 90°, and the little finger is the most flexible
sellar joint, exhibiting two degrees of freedom: flexion and (at about 95°), while the index finger is the least flexible (at about
extension, and abduction and adduction. The CMC joints of the 70°).16 The extension varies widely among individuals. For PIP and
fingers are classified as plane joints with one degree of freedom, DIP joints, flexion of about 110° and 90° occurs. Extension beyond
while the fifth CMC joint is often classified as a semi-saddle joint the zero position is regularly observed and depends largely on the
with conjunctional rotation.9 The next joint of each finger links the ligamentous laxity.
metacarpal bone to the proximal phalanx at the
metacarpophalangeal (MCP) joint. MCP joints are classified as 2.1.2 Muscles
ellipsoidal or condylar joints with two degrees of freedom,10 which Dexterous movements of the hand are accomplished by the
again permit flexion, extension, abduction, and adduction coordinated action of both the extrinsic and intrinsic musculature.
movements. In MCP joints, the metacarpal heads fit into shallow The extrinsic muscles originate from the arm and forearm, and they
cavities at the base of the proximal phalanges.11 The proximal are responsible for flexion and extension of the digits. The intrinsic
interphalangeal (PIP) and distal interphalangeal (DIP) joints are muscles are located entirely within the hand, and they permit the
found between the phalanges of the fingers; the thumb has only one independent action of each digit.17 There are nine extrinsic muscles,
INTERNATIONAL JOURNAL OF PRECISION ENGINEERING AND MANUFACTURING Vol. 13, No. 5 MAY 2012 / 809

(a) Direct matching of joint centers22 (b) Linkage for remote center of rotation23 (c) Redundant linkage structure24

(d) Tendon-driven mechanism25 (e) Bending actuator attached to the joint26 (f) Serial linkage attached to distal segment27
Fig. 2 Mechanisms for matching the center of rotation or eliminating the need for precise alignment

and three muscles among them - the flexor digitorum superficialis, metacarpal head and attaches to the volar plate, while the collateral
the flexor digitorum profundus, and the flexor pollicis longus - portion arises from the metacarpal head and attaches to the base of
contribute to finger flexion. Five extrinsic muscles contribute to the the phalanx. In contrast, the PIP and DIP joint collateral ligaments
extension of the fingers, while one extrinsic muscle (abductor attach completely to the bones. The collateral ligaments of the PIP
pollicis longus) contributes to the abduction of the thumb. The and DIP joints are concentrically placed and are of equal length;20,21
dorsal interossei (DI) and palmar interossei (PI) are groups of therefore, these ligaments are maximally taut throughout their range
muscles arising between the metacarpals and attached to the base of of motion.
the proximal phalanges or to the extensor assembly. The interossei
flex the MCP joint and extend the PIP and DIP joints. They are also 2.2 Requirements of the Hand Exoskeleton
effective abductors and adductors, and produce some rotations of One of the most important requirements of any device that
the MCP joint. Because of this interaction between the extrinsic and interacts with humans is safety. Because the exoskeleton devices
intrinsic musculature, the actions of the PIP and DIP joints are move under close contact conditions with the wearer, any
functionally coupled. malfunction can be seriously harmful to the user. Mechanical
designs should therefore consider the possibilities of unpredicted
2.1.3 Tendons and Ligaments erroneous operation of the device controller when the device is
As a digit moves, each tendon slides a certain distance. This actively actuated. Limits to the range of motion can be set using a
excursion takes place simultaneously in the flexor and extensor mechanical stopper or corresponding structural designs so that the
tendons.18 The relationships between the excursions of the finger exoskeleton cannot force the wearer’s body to move in an excessive
tendons and the angular displacements of the MCP, PIP, and DIP range of motion.
joints have been reported to be both linear and nonlinear.19 The The coincidence of the center of rotation is a primary concern in
excursions are larger in the more proximal joints. Also, the the mechanical design of hand exoskeletons. When the user wears a
excursion of the flexor tendons is larger than that of the extensor hand exoskeleton with rigid linkages, the linkage structure should
tendons, and the excursion of the extrinsic muscle tendons is larger be designed to have a center of rotation that coincides with the
than that of the intrinsic tendons. rotational axis of the human body joint. Otherwise, the difference in
There are a number of important extracapsular and capsular the rotational axes may cause a collision between the user’s hand
ligaments that support and stabilize the hand. The most important and the device, resulting in damage to the user’s hand.
extracapsular ligament is the transverse intermetacarpal ligament The most intuitive method is to build the exoskeleton’s center
(TIML). It attaches to and runs between the volar plates at the level of rotation to coincide with that of the wearer.22 However, this
of the metacarpal heads across the entire width of the hand. The requires an additional space to locate the mechanism at the side of
capsular collateral ligaments provide important joint stability to all the finger, making it difficult to build a multi-fingered structure.
of the finger and thumb joints. The MCP joint ligaments have dual Otherwise, a remote center of rotation can be adopted. There are
attachments: bony and glenoid. The glenoid portion arises from the various applicable mechanisms for the remote center of rotation for
 810 / MAY 2012 INTERNATIONAL JOURNAL OF PRECISION ENGINEERING AND MANUFACTURING Vol. 13, No. 5

Fig. 3 Classification of hand exoskeletons according to the various criteria

this purpose.23,28,29 However, the consideration of the coincidence exoskeletons provide exercise for the patients to help recovering
of the rotational axis can be disregarded when a flexible or motor function of hand. The rehabilitation exercise can be either
underactuated structure is adopted. For example, a linkage structure passive movement driven by the exoskeleton or active movement
with redundant degrees of freedom can be used.24 In this against the resistive force given by the exoskeleton. Therefore the
mechanism, the number of DOFs of the linkage structure use of sensors and actuators is not mandatory but depends on the
connecting the adjacent finger segments is 2 while that of human specific functions that are needed. On the other hand, the assistive
finger IP joint is only 1. The redundancy is eliminated by the exoskeletons acquire the user’s motion intention and assist the user
constraints given when attaching the device to the user’s hand. A performing the action. This functionality makes it necessary to be
tendon-driven mechanism mimicking the actuation of the actual equipped with sensors and actuators.
human hand can also be used for the actuation of the hand The hand exoskeletons can be classified using various criteria,
exoskeleton.25,30 Soft pneumatic actuators directly attached to the such as actuator type, power transmission method, degrees of
joint of a glove work in the same way.26 In these cases, where the freedom (DOF), intention sensing method, and control method.
flexible or underactuated structure is adopted, the wearer’s hand According to these criteria, hand exoskeletons can be classified as
actually provides a skeletal structure for the motion of the shown in Fig. 3. Among them, the type of actuator is selected as a
exoskeleton device. In addition, a serial linkage mechanism which major criterion for classification in this paper. Table 1 shows the
is attached only to the distal segment of the finger also does not passive exoskeleton. Table 2 and Table 3 show the rehabilitation
need the alignment of joint axis.27 Fig. 2 shows the mechanisms exoskeletons driven by electric actuators and pneumatic actuators,
described for matching the center of rotation or bypassing the respectively. In the same manner, Table 4, Table 5, and Table 6
problem. show the assistive exoskeletons driven by electric actuators,
Also, especially for the exoskeletons for assistance applications, pneumatic actuators, and shape memory alloy, respectively.
building a lightweight exoskeleton device and supporting
components must be considered a high priority. The power 3.1 Exoskeletons for Rehabilitation
transmission method and actuation mechanism must also be 3.1.1 Driven by Passive Actuator
considered with the structure as dominant factors in the design. 3.1.1.1 HandSOME31 (Fig. 4(a))
In addition to the factors described, the method for sensing the The Hand Spring Operated Movement Enhancer (HandSOME)
user’s intended motion is also a critical consideration and is closely is a passively operated device for giving an extension moment to
coupled with the device design. This will be further discussed later the finger joints so that it compensates for the finger flexor
in the paper in a dedicated section for intention sensing methods. hypertonia caused by a stroke. It is designed to follow the normal
kinematic trajectory of the hand during pinch-pad grasping,
providing an extension torque profile that best compensates for the
3. Review of Hand Exoskeletons finger flexor hypertonia. A 4 bar linkage mechanism was designed
for the thumb and finger parts to coordinate the natural grasping
Several research groups have developed hand exoskeletons for motion. The attachment point of the spring can be changed to adjust
rehabilitation and assistance applications. The rehabilitation the torque profile.
INTERNATIONAL JOURNAL OF PRECISION ENGINEERING AND MANUFACTURING Vol. 13, No. 5 MAY 2012 / 811

Table 1 Rehabilitation exoskeleton driven by passive actuator

Reference Force transmission DOF Note
HandSOME (Brokaw et al.)31 Linkage 1 Exert extension torque for compensating finger flexor hypertonia

Table 2 Rehabilitation exoskeletons driven by electric actuators

Reference Force transmission Active DOF Intention sensing method Note
WaveFlex (Otto Bock)32 Linkage 1 CPM
Kinetec Maestra Portable Hand
Linkage 1 CPM
CPM (Patterson Medical)33
Mulas et al.34 Cable 2 EMG Active control
Tong et al.35 Linear actuator 5 EMG CPM / Active motion
HEXOSYS (Iqbal et al.)36 Linkage 2 Underactuated
HEXORR (Schabowsky et al.)37 Linkage 2 Torque sensor CPM / Active motion
HANDEXOS (Chiri et al.)38,39 Cable, crank-slider 5 Underactuated
Wege et al.24,40 Cable 20 EMG electrode Active motion
Ueki et al.41 Linkage 18 Joint angles of healthy hand Self-motion control
iHandRehab (Li et al.)42 Cable 8 Force sensor CPM / Active motion
Sarakglou et al.43 Cable 7 Virtual reality exerciser
AFX (Jones et al.)44 Cable, linkage 3
IntelliArm (Ren et al.)45 Linkage 1 for hand Passive / assistive

Table 3 Rehabilitation exoskeletons driven by pneumatic actuators

Reference Force transmission Active DOF Intention sensing method Note
Hand Mentor (Kinetic Muscles)46 Linkage 1 Passive / assistive
HWARD (Takahashi et al.)47 Linkage 3 Assistive
Wu et al.48 Cable, linkage 2 Force sensor Assistive

Table 4 Assistive exoskeletons driven by electric actuators

Reference Force transmission Active DOF Intention sensing method Note
Underactuated
Martinez et al.49,50 Cable 3 FSR
Passive extension
51
OHAE (Baker et al.) Cable 3 FSR Underactuated
Hasegawa et al.52,53 Cable 11 EMG Finger tracking for back-drivability
Underactuated
In et al.30 Cable attached to glove 1 EMG
Passive extension
Underactuated
In et al.25 Cable attached to glove 1 Passive extension,
Differential mechanism
Shields et al.54 Cable, linkage 3 Force sensors Passive extension
SkilMate (Yamada et al.)55 Steel belt 3 Joint angle Equipped with tactile sensor at fingertip
Benjuya et al.56 Flexible shaft 1 EMG

Table 5 Assistive exoskeletons driven by pneumatic actuators

Reference Force transmission Active DOF Intention sensing method Note
DiCicco et al.57,58 Cable, linkage 2 EMG Passive extension
Underactuated
Sasaki et al.59 Directly attached to glove 6 Expiration switch or tactile sensor
Passive extension
26
Kadowaki et al. Directly attached to glove 6 Flexion angle or EMG Underactuated
60 Directly attached beneath the Underactuated
Tadano et al. 5 Force sensor
finger linkage Passive extension
61
Takagi et al. Linkage 3 Bending sensor Passive extension
Toya et al.62 Directly attached to glove 4 Estimate from movement pattern Passive extension
Moromugi et al.63 Linkage 1 Muscle hardness sensor

Table 6 Assistive exoskeleton driven by shape memory alloy

Reference Force transmission Active DOF Intention sensing method Note
Makaran et al.64 Linkage 1 Sip-and-puff switch or EMG Passive extension
 812 / MAY 2012 INTERNATIONAL JOURNAL OF PRECISION ENGINEERING AND MANUFACTURING Vol. 13, No. 5

3.1.2.4 Tong et al.35

Tong et al. presented a hand exoskeleton which consists of 5
finger assemblies where each finger has 1 active DOF actuated by a
linear actuator, causing coupled movement of the MCP and PIP
joints.
The device has 4 modes of operation: 1) CPM, 2) EMG-
triggered motion, 3) continuous EMG-driven motion, and 4) free-
(a) HandSOME31 (b) HandEXOS38,39 running. In the second mode, the device starts flexion or
extension motion when the corresponding EMG signal exceeds a
certain threshold. In the third mode, the movement continues as
long as the user’s effort exists. The fourth mode selects flexion or
extension of the device according to a comparison of the EMG
signals from the two muscles that represent flexion and
extension.

(c) Wege et al.24,40 (d) Ueki et al.41 3.1.2.5 HEXOSYS36

Fig. 4 Some of the hand exoskeletons for rehabilitation Iqbal et al. proposed the Hand EXOskeleton SYStem
(HEXOSYS), which actuates 2 fingers for rehabilitation. Each
3.1.2 Driven by Electric Actuator finger is driven by using an underactuated linkage driven by an
3.1.2.1 WaveFlex32 electric motor. The linkage structure adopted in this device is a
The WaveFlex (Otto Bock, Germany) is a commercial three-link planar underactuated mechanism having a single
continuous passive movement (CPM) device for physical therapy of attachment point. A custom-made force sensor is integrated into the
the hand. An electric motor is used for actuation. This device connecting link.
achieves a full range of motion (ROM) of flexion and extension
using a drive bar and finger attachments to assist the fingers 3.1.2.6 HEXORR37
through a natural path for a grasping motion. The WaveFlex is The Hand Exoskeleton Rehabilitation Robot (HEXORR)
portable and lightweight, enabling it to be worn for extended developed by Schabowsky et al. consists of two modular
periods of time, and is adjustable for different finger lengths using components; one is for the fingers, while the other is for the thumb.
the attached finger clips. The WaveFlex is also able to measure the The finger module is built with a four-bar linkage that is capable of
interaction force. When the interaction force exceeds a certain providing coupled rotations of the MCP and PIP joints. Each
threshold during motion, the ‘reverse-on-load’ function controls the module is driven by an electric motor and the user’s movement
device to move in the reverse direction to prevent overloading of volition is sensed using a torque sensor.
the user’s fingers. The user can also use this device to exercise the This device has three modes of operation: 1) CPM, 2) active
thumb. However, it is not possible to move the thumb unassisted movement, and 3) active force assisted movement. In the
simultaneously with the other fingers. second mode, the device compensates for the weight and friction of
the device itself, while rejecting unintentional movement
3.1.2.2 Kinetec Maestra Portable Hand CPM33 commands. The third mode provides assistance for extension
The Kinetec Maestra Portable Hand CPM (Patterson Medical, movements.
USA) is a commercial CPM device for hand rehabilitation. It
incorporates a bilateral Alumafoam splint for attachment of the 3.1.2.7 HANDEXOS38,39 (Fig. 4(b))
device to the user’s forearm. Flexion and extension movements are The hand exoskeleton developed by Chiri et al. has 5
made via a drive bar to which the 4 fingers other than the thumb are independent modules for the fingers. Each module is composed of 3
connected together. The drive bar is actuated using an electric motor. links for the phalanges, where the center of rotation of each
The device can provide hyperextension and full flexion for the connection is matched with the corresponding joint of the human
fingers, but thumb movement is not involved. finger. The flexion and extension of the MCP joint is driven by a
slider-crank-like mechanism, while the PIP and DIP joints are
3.1.2.3 Mulas et al.34 driven by Bowden cable transmissions. The 3 joints of each finger
A device developed by Mulas et al. is actuated using two are underactuated because they are driven using a single actuator
electric motors that drive wires to flex the thumb and the other unit.
fingers. Extensions are performed using springs. Unlike the CPM For the finger module, 3 force sensors are mounted on the
devices, this device is controlled based on an electromyography surface of the inner side of each of the three palmar shells to
(EMG) signal to start the movements according to the user’s sense the interaction force. The linear slider for MCP rotation is
volition. When the EMG signal exceeds a certain threshold, the equipped with strain gauges to measure the force transmitted by
flexion movement is initiated. the driving cable.
INTERNATIONAL JOURNAL OF PRECISION ENGINEERING AND MANUFACTURING Vol. 13, No. 5 MAY 2012 / 813

3.1.2.8 Wege et al.24,40 (Fig. 4(c)) This device can be used for virtual reality based physical
The hand exoskeleton developed by Wege et al. actuates each exercise, where a patient performs physical and occupational
joint via a Bowden cable driven by an electric motor. Bidirectional therapy exercises by interacting with a number of virtual simulated
movement is supported by the use of two pull cables for each joint, exercises that are designed in a game-like fashion.
diverted by a pulley on both ends. Only one motor is used for each
joint, which introduces some slackness when compared to a 3.1.2.12 AFX44
solution using one motor for each direction. The motion is applied Jones et al. proposed the Actuated Finger Exoskeleton (AFX),
through a leverage construction on each finger attachment. which has 3 active DOF for the index finger joints: the MCP (1
This device is controlled by EMG signals. Each finger rests in DOF), PIP (1 DOF), and DIP (1 DOF), actuated by a cable
its relaxed position when no muscle activation is measured. mechanism driven by electric motors. The three rotational joints of
Depending on the muscle activation, a linear force is calculated and the exoskeleton are aligned with the flexion/extension axes of each
the fingers are moved as if acting against a constant friction. The joint of the user. The exoskeleton structure is therefore located at
movements of the MCP, PIP, and DIP joints are performed in a the side of the finger. This device is capable of operating in position
coupled motion. control mode or torque control mode.

3.1.2.9 Ueki et al.41 (Fig. 4(d)) 3.1.2.13 IntelliArm45

Ueki et al. proposed a hand exoskeleton for hemiplegic patients. Ren et al. developed a whole arm exoskeleton with a hand part
The device is capable of 18 DOF motions: 3 DOF for each finger, 4 actuated by four bar linkages and electric motors. One active DOF
DOF for the thumb, and 2 DOF for the wrist. For each finger, 3 was designed to drive the hand to open/grasp at the MCP and thumb
electric motors assist the flexion/extension of the MCP and PIP joints in a synchronized opening/closing motion of the hand. An
joints and the abduction/adduction of the MCP joint. For the thumb, electric motor is used to provide hand opening and closing training.
there are 3 motors for flexion/extension and one for opposition. The Passive movement and active assistive exercise are provided
wrist motion is performed using 2 motors. with this device. The active assistive exercise mode can improve
The device is controlled to reproduce the movements of a voluntary neuromuscular control by using games with a gripping
healthy arm. A data glove is used to measure the joint angles of a task.
healthy arm and the hand exoskeleton mimics the measured joint
motion. 3.1.3 Driven by Pneumatic Actuator
3.1.3.1 Hand mentor46
3.1.2.10 iHandRehab42 The hand mentor is a commercial hand rehabilitation therapy
The iHandRehab proposed by Li et al. aims to satisfy the system produced by Kinetic Muscles Inc. (USA). It is a 1 DOF
requirements for both active and passive movements for hand device that provides a controlled resistive force to the hand and
rehabilitation. This device has finger modules for the index finger wrist. The applied force can oppose flexion or assist extension of
and thumb. The index finger part consists of the MCP (2 DOF), PIP the hand. It incorporates sensors that monitor the position of the
(1 DOF), and DIP (1 DOF) modules, and the thumb consists of the wrist and fingers during flexion/extension motions, as along with
CMC (2 DOF), MP (1 DOF), and IP (1 DOF) modules. All actuated force sensors to measure the force applied to the hand by the
joints are driven by cable transmissions. To realize bidirectional compliant air muscle actuator. The device incorporates surface
movement, two cables were used for each joint motion. EMG recording electrodes in contact with the patient’s muscles and
This device can operate in passive, active, and assisted modes. an EMG level display.
In the active modes, a force control scheme is implemented to exert
a resistive force on the user’s fingers. Force sensors are used to 3.1.3.2 HWARD47
measure the interaction forces at the fingertips. The assisted mode The Hand Wrist Assistive Rehabilitation Device (HWARD)
switches from the active mode to the passive mode during the developed by Takahashi et al. is a 3 DOF (1 for fingers, 1 for thumb,
exercise. and 1 for wrist) pneumatically actuated system that exercises
flexion and extension of the hand as well as wrist movement. The
3.1.2.11 Sarakoglou et al.43 device can simultaneously flex and extend the fingers, including the
Sarakoglou et al. developed a hand exoskeleton to provide thumb, about the MCP joint. Wrist flexion and extension is also
physiotherapy regimes in an interactive virtual environment. This performed. This device can assist with grasping and releasing
device provides facilities for hand motion tracking, recording and movements while simultaneously allowing the user to feel real
analysis as well as the ability to execute both occupational and objects during therapy. Three double-acting cylinders are used to
physical therapy exercises. It provides 7 active DOF: 2 for each drive the device.
finger except for the thumb (1 DOF). The device is actuated by
pulling cables driven by electric motors located at the motor site. To 3.1.3.3 Wu et al.48
measure the interaction forces, force sensors are also installed at the Wu et al. developed a hand exoskeleton with 2 active DOF
motor site. (flexion/extension of the MCP and PIP joints of the fingers,
 814 / MAY 2012 INTERNATIONAL JOURNAL OF PRECISION ENGINEERING AND MANUFACTURING Vol. 13, No. 5

excluding the thumb). This device provides the assistive forces

required for finger training. To enable bidirectional movement at a
finger joint with a pneumatic muscle, a PM-TS actuator consisting
of a pneumatic muscle and a torsion spring is applied. In this
configuration, the torsion spring provides the extension of the
pneumatic muscle.
The purpose of the control scheme used in this device is to (a) Hasegawa et al.52,53 (b) In et al.25
provide controllable, quantifiable assistance specific to some
particular patients by adapting the level of assistance provided.

3.2 Exoskeletons for Assistance

Various works have been conducted for applications in hand
function assistance. The purpose of most of these devices is to help
the disabled. However, some of the devices were developed to help
astronauts, because moving fingers while wearing a space suit glove (c) Shields et al.54 (d) DiCicco et al.57,58
is difficult because of the stiffness of the glove itself and the
pressure difference.

3.2.1 Driven by Electric Actuator

3.2.1.1 Martinez et al.49,50
At the College of New Jersey, a power-assisted exoskeleton has
been designed to help the pinching and grasping motion of people (e) Kadowaki et al.26 (f) Tadano et al.60
with decreased hand functionality caused by disease. Martinez et al. Fig. 5 Some of the hand exoskeletons for assistance
designed an under-actuated cable-driven exoskeleton with active
flexion and passive extension mechanisms. There are three actuated finger motion driven by tendons, there is a difference in that their
fingers: the thumb, index and middle fingers. The middle finger device controls each joint independently. This method is used to
motion acts in conjunction with that of the ring and small fingers. simulate the compliance variation of a human finger according to
For each finger, flexion is performed using a linear actuator, while the grasping force exerted to maintain grasping stability.
extension is performed by a spring. Aluminum bands are located at The authors proposed a ‘dual sensing system’ and a ‘bioelectric
the circumferences of the phalanges, forming a linkage with potential-based switching control algorithm’ to enable small
connecting structures between the bands. Force sensing resistors resistance to movement while providing force augmentation only
(FSR) installed inside the actuated fingers measure the flexion when the user exerts a relatively large grasping force. The finger
forces for control of the device. joint angles and the bioelectric potential are measured to control the
device. The grasping force is estimated from the bioelectric
3.2.1.2 Orthotic Hand-Assistive Exoskeleton (OHAE)51 potential measured by surface electrodes on the lumbrical muscles.
Baker et al. introduced a project to develop a hand exoskeleton When the estimated grasping force is below a certain threshold,
with an active extension capability, unlike the previous exoskeleton meaning that the force assistance is not required, the device controls
designs49,50 described above that used springs to extend the fingers. the motors to keep the wires slightly relaxed, regardless of the
This device has three actuated fingers: the thumb, index and middle finger posture. The motor control commands are generated by
fingers, driven by cables attached to a glove. Aluminum bands and calculation of the required wire lengths based on the joint angles
carbon fiber rods sewn into the glove build a skeletal structure for measured from the exoskeleton. This behavior results in low
finger movement. There is a linear actuator for each finger, which resistance during unassisted finger movement. However, if the
pulls the cable in bidirectional motion to flex and extend the finger. estimated grasping force becomes significantly large, indicating that
The motion intention of the user is sensed by two force-sensing the user needs force assistance, the control mode of the exoskeleton
resistors (FSR) attached at the dorsal and ventral sides of the distal is switched to the other mode, which controls the grasping force.
link of each actuated finger. The FSRs are intended to measure the Using this mode, assistance is given to the index finger while the
contact forces caused by the user’s finger movement. thumb maintains its current posture.

3.2.1.3 Hasegawa et al.52,53 (Fig. 5(a)) 3.2.1.4 In et al.30

Hasegawa et al. have developed an exoskeleton to assist with In et al. proposed a glove-type hand exoskeleton to assist
hand and wrist functions. The device has a total of 11 active DOF: disabled people. This device adopts an underactuated cable-driven
three for the index finger, three for the middle-ring-small finger mechanism attached to a glove. Because there is no rigid linkage,
combination, two for the thumb, and three for the wrist. Although the wearer’s hand becomes the linkage structure for operation of the
the authors adopted a cable-driven mechanism mimicking human exoskeleton. A cable exerts a flexion force on each finger, while the
INTERNATIONAL JOURNAL OF PRECISION ENGINEERING AND MANUFACTURING Vol. 13, No. 5 MAY 2012 / 815

extension force is provided passively by a spring. All of the on/off manner with two threshold levels that classify the operation
actuated fingers are driven by a single motor. However, the tendon modes into flexion, stop, and extension modes.
excursions which occur during the finger movements are different
for each finger because of the differences in the moment arms. 3.2.1.7 SkilMate55
There are therefore stacked pulleys with different diameters at the Yamada et al. proposed a design for a powered hand assistance
output shaft of the motor, providing suitable amounts of tendon device for space suit gloves. Three fingers are actuated using the
excursion for each finger. An electromyography (EMG) signal is device: the thumb, index, and middle fingers. The largest joint for
used to control the device in a simple on-off manner. The device each finger is actuated by an ultrasonic motor to flex or extend the
exerts a flexion force when the EMG signal exceeds a predefined joint. The device is composed of inner and outer parts
threshold. corresponding to master and slave devices, respectively. The outer
part is controlled to follow the motion of the inner part. The joint
3.2.1.5 In et al.25 (Fig. 5(b)) angle of each actuated finger is measured using an encoder attached
After the preceding work30 described above, In et al. developed to the inner part.
another hand exoskeleton, adopting a differential mechanism for Because of the importance of tactile information in
multi-finger underactuation to substitute for the stacked pulleys manipulation, this device is designed to be equipped with tactile
with different diameters that were used in the previous model. Like sensors and tactile display elements to provide the wearer with
its predecessor, this device uses the user’s own hand as a supporting tactile information in the form of vibration.
structure for finger movement, because there are no rigid linkages.
The flexion motions of the three actuated fingers are performed 3.2.1.8 Benjuya et al.56
using a motor, and the extension motions are performed using Benjuya et al. developed a myoelectric hand orthosis for spinal
extension springs. cord injury patients at the C5-6 level. This device has one actuated
The differential mechanism enables the device to grasp an DOF at the MCP joint for flexion/extension of the coupled index
object with a three-dimensional surface securely with only one and middle fingers. A DC motor is located on a forearm band,
actuator by adjusting the movement of the fingers. The key parts of transmitting power to the fingers through a flexible shaft. The
the proposed differential mechanism are U-shaped tubes located at flexible shaft has a worm gear at the distal end so that the shaft
the fingertips and between the fingers. The tubes at the fingertips rotation drives a spur gear of a finger piece, to which the index and
move with the fingers, while the tube between the fingers maintains middle fingers are tied. The pinching force is controlled in a manner
its position. When a spooler attached to a motor pulls the cable for proportional to the amplitude of the EMG signal from the forearm.
finger flexion, the total exposed length of the flexor cable is
shortened, and this causes the flexion of the fingers. When there is 3.2.2 Driven by Pneumatic Actuator
no external resistance, the actuated fingers are flexed almost evenly. 3.2.2.1 DiCicco et al.57,58 (Fig. 5(d))
However, if one finger is blocked by an obstacle, the U-shaped tube DiCicco et al. developed an orthotic hand exoskeleton for
of the obstructed finger cannot move any further. On the other hand, quadriplegic patients with C5/C6 injuries. With this device, a
shortening of the flexor cable results in faster flexion of an pinching motion is performed by the index finger while the thumb
unobstructed finger. is fixed in an opposed posture. This system has 2 active DOF for
the index finger: one for MCP flexion/extension, and the other for
3.2.1.6 Shields et al.54 (Fig. 5(c)) coupled PIP/DIP flexion/extension. The flexion of the PIP and DIP
Space suits and gloves are stiffened by the pressure difference joints is controlled using a cable located at the volar side of each
when they are exposed to the vacuum of space during finger band. These cables are pulled by a pneumatic cylinder acting
extravehicular activities (EVA). Because it is difficult for astronauts in compression. The flexion of the MCP joint is performed by a
to move against this stiffness, space suits have caused reduced linkage mechanism driven by a pneumatic cylinder acting in
dexterity and increased fatigue. To overcome this problem, some extension. Pressurized air is supplied to the pneumatic cylinders
devices have been developed. simultaneously. For extension of the joints, springs are mounted at
Shields et al. proposed a hand exoskeleton for an EVA glove. It the joints to exert a passive extension force.
has three actuated fingers (index, middle, ring-small), with one Three control strategies are applied for control of the device.
DOF for each finger. The links for each finger form four-bar First, a binary control algorithm with a simple on/off method based
mechanisms to allow the joints to rotate about remote centers that on the EMG signal acquired from the biceps of the contralateral
are coincident with the joints of the wearer’s fingers. The motions arm can be used. With this control mode, the finger is flexed when
of the two joints for each finger are coupled together. This device the signal level from the contralateral biceps exceeds a certain
exerts a flexion force generated by motors via a cable-driven cam threshold. The flexed posture is maintained while the signal level
mechanism, while the extension is performed using a passive force remains above the threshold. Second, a method which controls the
provided by the stiffness of the space suit glove. The user’s air pressure continuously relative to the measured EMG signal from
intention to flex the glove is sensed by force sensors mounted inside the contralateral biceps is applied. Finally, a natural reach and pinch
each fingertip. The control of the device is performed in a simple algorithm which uses the EMG signal from the ipsilateral biceps is
 816 / MAY 2012 INTERNATIONAL JOURNAL OF PRECISION ENGINEERING AND MANUFACTURING Vol. 13, No. 5

used. With the third control mode, the user does not have to pneumatic artificial rubber muscles. Although the device has a total
concentrate on straining their contralateral arm to control the device. of 10 DOF comprising 2 DOF for each finger, they are
underactuated, with one active DOF for each finger. A contracting
3.2.2.2 Sasaki et al.59 pneumatic rubber muscle is attached under a bi-articular linkage
Sakaki et al. developed a wearable power assisted device for mechanism for each finger for flexion.
grasping functions. The device has five fingers actuated by At the fingertip part of each finger, a balloon sensor is installed
pneumatic rubber muscles. Each pneumatic muscle is attached to sense pressure exerted by the user. The pressure values sensed by
directly to the glove, eliminating the usage of a linkage structure. the balloon sensors are applied to grasping force control of the
Each finger, except for the thumb, has one active DOF for device. The device amplifies the grasping force in proportion to the
flexion/extension, while the thumb has 2 active DOF for sensed pressure.
flexion/extension and for opposing motion.
A curved type rubber muscle is used for the flexion of each 3.2.2.5 Takagi et al.61
finger, including the thumb, while two linear type rubber muscles Takagi et al. developed a grip aid system using pneumatic
are used for the opposing movement of the thumb. The curved type cylinders. It has three actuated fingers: the thumb, index, and
tuber muscle is composed of a lengthwise expandable rubber tube middle fingers. Each finger is equipped with a pneumatic cylinder
with an inelastic fiber tape attached to the side of it. Pressurization at the dorsal part of the finger so that extension of the pneumatic
of the rubber tube makes the rubber muscle bend. The difference cylinder causes the flexion of the corresponding finger. The linkage
between the linear type rubber muscle and the curved type rubber mechanisms for the index and middle fingers cause coupled MCP-
muscle is the absence of the fiber tape. Therefore, when pressurized, PIP joint motion.
the linear type rubber muscle is extended in the axial direction. A bending sensor attached to the small finger measures the
One of the operating methods for the device is on/off control flexion angle of the small finger. The sensed bending angle can be
using an expiration switch. When the pressure provided by the used for control of the device.
user’s mouth exceeds a certain threshold, the device is activated for
grasping. The other operating method is contact force control using 3.2.2.6 Toya et al.62
a tactile sensor installed at the index fingertip. The pressure of the Toya et al. developed a power-assisted glove which is
supplied air is feedback-controlled by this method. controlled based on the estimated grasping intention extracted from
the initial movement patterns of the finger joint angles. The device
3.2.2.3 Kadowaki et al.26 (Fig. 5(e)) assists all 5 fingers. Each finger has 2 active DOF, apart from the
Kadowaki et al. developed a power-assisted glove for those who thumb, which has one active DOF. However, the MCP joints of the
have a weak hand grasping force. The actuated DOF are the same index, middle, ring, and small fingers are actuated together. The PIP
as for its predecessor, described above.59 This device also adopted joints of the index and middle fingers are also actuated
pneumatic rubber muscles as actuators. The differences between simultaneously. In the same manner, the PIP joints of the ring and
this device and that of the former work are the types of pneumatic small fingers move together. Only the actuation of the thumb is
muscles used and the operating method. isolated. Therefore, the actual number of actuated DOF is 4. The
Two types of pneumatic rubber muscles are applied: one is a actuation is performed using pneumatic soft actuators that bend
sheet-like curved rubber muscle, and the other is a spiral rubber when pressurized air is supplied.
muscle. The former has a role in the flexion of each finger while the Unlike other hand assisting exoskeletons, this device performs a
latter makes the opposing motion of the thumb. Because the sheet- predefined motion from 3 grasping motions according to a
like curved rubber muscle has two lengthwise expandable elements classification result from analysis of the initial motion of the fingers.
located in parallel, the bending direction can be controlled by The three principal grasping motions applied are a power grip, a
selecting the element to be pressurized. Both the extension and the precision grip, and a tip pinch. For control of the device, four angle
flexion are therefore actively performed. The spiral rubber muscle sensors are installed in some of the joints. The angle sensor
consists of an expandable rubber tube and a cloth which is locations are determined based on the analysis of the initial
stretchable in the oblique direction. This makes the spiral muscle movement patterns of the finger joint angles for each grasping
twist when it is pressurized. mode. A pattern classification method is applied to the measured
The glove is controlled by means of finger posture, measured angles to distinguish the movement patterns and to predict the
using a data glove or an EMG signal acquired from the forearm grasping mode.
muscles. With a data glove equipped with bend sensors, the device
can be controlled using the motion of the glove. For the EMG-based 3.2.2.7 Moromugi et al.63
control case, the grasping motion commences when the signal level Moromugi et al. developed a hand exoskeleton actuated by a
exceeds a certain threshold. pneumatic cylinder for assisting with grip force. The device has an
actuated index finger with 3 links, where the links are connected
3.2.2.4 Tadano et al.60 (Fig. 5(f)) together by sublinks so that the motion of the pneumatic cylinder
Tadano et al. developed a hand exoskeleton actuated by causes synchronized motion at the joints. On the extension of the
INTERNATIONAL JOURNAL OF PRECISION ENGINEERING AND MANUFACTURING Vol. 13, No. 5 MAY 2012 / 817

cylinder, the exoskeleton performs a gripping motion toward the

fixed thumb. The user’s intention of motion is sensed using a
muscle hardness sensor attached to the forearm. The muscle
hardness sensor measures pressure while providing a mechanical
indentation on the skin. When the muscle under the sensor is
activated, the hardness increment of the muscle causes elevation of
the measured pressure.

3.2.3 Driven by Shape Memory Alloy

3.2.3.1 SMART Wrist-Hand Orthosis64
Makaran et al. developed an exoskeleton type hand orthosis to
help the grasping function of quadriplegic patients. The device has Fig. 6 Classification of electric motors
one actuated finger which rotates around the MCP joint axis. A
shape memory alloy (SMA) actuator is used as an actuator for the with electric motors, and they showed a long list of DC or BLDC
flexion of the finger. Extension is performed by a spring. Because motors. The DC motor has been extensively used because of the
the SMA used has high electric resistance, heat generation by simple structure of the motor itself, as well as that of its electronic
passing an electric current through it is a possible method of drive; however, it requires regular maintenance because of the
operating the SMA actuator. mechanical contact between the brush and commutator. The BLDC
The device is controlled by using a sip-and-puff switch or an motor not only requires no regular maintenance but also has the
EMG signal. They can be used as commands for on/off operation advantage of high speed driving because it uses an electronic
with appropriately defined thresholds. inverter instead of the brush and commutator. Also, a heavy
armature rotates in the DC motor while a light permanent magnet
rotates in the BLDC motor; the small inertia of the BLDC motor
4. Actuator Technologies therefore enables rapid acceleration and deceleration.66,67
To transmit the power of electric motors to each joint of the
Different types of actuators have been developed to actuate exoskeletons, transmission mechanisms such as cables, gears and
hand exoskeletons for assistive and rehabilitation purposes. In this linkages have been used.68
section, the conventional exoskeleton actuators (electric motor and
pneumatic actuator) and the smart material actuators (shape 4.2 Pneumatic Actuators
memory alloy and electroactive polymer) are introduced, and their Pneumatic actuators have been used in many exoskeleton
characteristics are briefly summarized. applications.68 The air compressor used to generate the compressed
air for pneumatic actuation is both bulky and noisy. The noise
4.1 Electric Motors problem can be overcome by using pre-compressed air storage.
Electric motors have been used successfully not only as However, the size of the pneumatic system cannot be easily reduced
exoskeleton actuators but also as prosthetic finger actuators, because of the air storage chamber volume. Pneumatic actuators
because they are easily available, reliable and easy to control. They therefore must be used for systems with lower mobility or their
can be categorized into DC motors and AC motors according to bulky parts must be placed in the user’s carrying case, such as in a
their electric power sources. The AC motor can further be classified wheelchair.58 Cylinders and pneumatic artificial muscles are widely
as shown in Fig. 6. Synchronous motors using permanent magnets used to transmit the power of compressed air into the
are classified into brushless AC (BLAC) motors and brushless DC exoskeletons.68
(BLDC) motors, depending on the shape of the back electromotive The McKibben type pneumatic artificial muscle is made of a
force. More specifically, BLAC and BLDC motors have sinusoidal rubber inner tube covered with a shell braided by helical weaving.
and trapezoidal back electromotive force shapes, respectively. The When the inner tube is pressurized, the muscle inflates and
BLAC motor system is generally more expensive than the BLDC contracts.69 Another commonly used form of pneumatic artificial
motor system.65-67 muscle is the bending type pneumatic muscle. Noritsugu et al.
The development of power electronics enables AC motors to be developed a pneumatic rubber muscle consisting of a rubber tube
widely used as actuators. The source of the field flux in a with a bellows sleeve.70 One side of the muscle was reinforced with
synchronous motor can be changed from an electrically excited fiber tape to generate a bending motion of the pneumatic muscle by
field winding to a permanent magnet by the use of a high- supplying compressed air. To replace the fiber reinforcement of the
performance reliable permanent magnet. The use of the permanent- bending type pneumatic muscle, Takashima et al. used a shape
magnet synchronous motor (PMSM) can increase the torque and memory polymer (SMP) with an elastic modulus that varied with its
power density with improved efficiency compared to that of a temperature.71 In this pneumatic muscle with SMP, the bending
synchronous motor with an electrically excited field winding.66 direction could be changed by varying the heating area of the
Gopura et al.68 summarized many upper limb exoskeletons actuated actuator.
 818 / MAY 2012 INTERNATIONAL JOURNAL OF PRECISION ENGINEERING AND MANUFACTURING Vol. 13, No. 5

structure and actuation mechanism of a dielectric elastomer. The

dielectric elastomer consists of a dielectric film with two surface
electrodes. When a high voltage is applied to the two electrodes, the
dielectric film becomes thinner, which results in its lateral
Fig. 7 Structure and bending mechanism of IPMCs. The positive expansion.73,77 Herr et al. introduced the application of a dielectric
and negative symbols represent cations and anions, and small elastomer to act as a bicep on a full size skeletal muscle.78 This
circles represent water molecules dielectric actuator needs a power transmission mechanism to be
used for a hand exoskeleton because it yields a linear motion.

4.4 Shape Memory Alloy Actuator

The shape memory alloy (SMA) actuator utilizes the shape
memory effect (SME), which indicates the property of recovering
the original shape upon heating to a critical temperature when it is
deformed in the low temperature phase.79 The materials that can be
used as SMA include Ni-Ti and Cu-Al-Ni, but several other
combinations exist. The SME occurs by the shift of crystalline
structure between two phases, martensite and austenite. It is in
martensite phase when the temperature is low. Heating above the
transition temperature makes it recover the original shape with
Fig. 8 Structure and actuation mechanism of dielectric elastomer returning to the austenite phase.80
Because of the unique property and the high power-to-weight
4.3 Electroactive Polymer Actuator ratio, they are being used for wide applications as both actuators
Though the electroactive polymers (EAPs) are not widely used and sensors. However, the high nonlinearity including hysteresis
as actuators for exoskeletons, they are attractive actuators because and saturation make the precise control of the SMA actuator
of their muscle-like nature, such as light weight, flexibility and low difficult.
power consumption. They can be classified into ionic type and
electronic type EAPs. The ionic EAP generates deformations such
as expansion, contraction or bending through movement of ions in 5. Intention Sensing Methods
response to voltage stimulations as low as 1-5 V. Ionic polymer-
metal composites (IPMCs), ionic polymer gels, conductive For assistive hand exoskeletons, accurate sensing of the user’s
polymers and carbon nanotubes are ionic EAPs. This type of EAP intended motion is a primary concern. For the purposes of
has the advantages of low drive voltage, large bending displacement controlling a device or ergonomic evaluation, there have been
and natural bi-directional actuation, along with the disadvantages of various methods for detection of motion intention. The applied
slow response and a relatively low actuation force.72 Fig. 7 shows techniques range from direct measurement of contact force to
the typical structure and actuation mechanism of IPMCs. IPMCs estimation of the exerted force from biomechanical signals.
consist of an ionic polymer membrane and two surface metal The methods mentioned below contain not only methods that
electrodes. When a low voltage is applied to the two electrodes have already been applied to hand exoskeletons, but also those that
(anode and cathode), cations in the polyelectrolyte move towards have not been applied yet. The latter methods have either been
the cathode; the cathode side therefore swells while the other side adopted in other interactive devices or have potential for usage in
shrinks, which results in the bending deformation.73 Bar-Cohen hand exoskeletons. In fact, the intention sensing methods can be
introduced a 4 finger gripper lifting a rock as a robotic application used as a general means of device control.
of IPMCs.74 Also, Deole et al. developed an IPMC microgripper to
manipulate micro-sized objects.75 This IPMC actuator does not 5.1 Force Sensing
require a power transmission mechanism for hand exoskeleton One of the most direct methods of sensing a user’s intention is
applications because it generates a natural bending motion like the to measure the force exerted by the user at the interface. This
aforementioned pneumatic muscle. method has been applied to several hand exoskeletons for assistance
In contrast to the ionic EAP, the electronic EAP is driven by an applications.49-51,54,59,60 The sensing is usually performed at the
electric field or by Coulomb forces. Dielectric elastomers, fingertip. Although it may obstruct the haptic sensation of the user’s
ferroelectric polymers and electrostrictive graft elastomers are types finger by preventing the finger from contacting an object which is
of electronic EAP. This type of EAP has the advantages of rapid to be manipulated, it is the most reliable method for control of the
response and a relatively large actuation force; however, it requires grasping force. Also, obstruction of the haptic sensation is not a
heavy components such as high voltage transformers and has problem for assistive devices for EVA gloves. For measurement of
potential problems related to safety issues and material breakdown the contact force, force sensing resistors (FSRs), pneumatic
because of the high actuation voltage.72,73,76 Fig. 8 shows the pressure sensors, and strain gauge sensors are predominantly used.
INTERNATIONAL JOURNAL OF PRECISION ENGINEERING AND MANUFACTURING Vol. 13, No. 5 MAY 2012 / 819

5.2 Motion Sensing acquisition of the sEMG signal from these muscles difficult. Also,
The bending angle of the finger can be used as an input signal many muscles contribute to the pinch force generation,90 causing
for a position controller to operate a hand exoskeleton.26,55,61,62 crosstalk of the signals from the active muscles.91,92
However, because the bending angle of the finger should be However, using sEMG has some difficulties:93 Because the
induced by the user’s motion, hand exoskeletons of this type electrical potentials measured by sEMG are very weak,
usually have a master-slave configuration. In this case, the master measurement requires careful electrode placement and excellent
device is closely attached to the user’s finger to measure the finger contact with the skin. The skin humidity and electrode location can
posture. The slave device, which is the assistive hand exoskeleton, also affect the measurement results greatly.
follows the posture of the master device when the movement of the
master device occurs. The hand exoskeleton can also be controlled 5.5 Muscle Hardness
by a finger that is not assisted by the device.61 The initial movement The contraction of a muscle causes an increment in the muscle
pattern of the user’s finger can also be a triggering command for hardness. Acquiring the muscle hardness by measuring the pressure
programmed grasping based on a pattern classification technique.62 under a certain skin deformation caused by a mechanical
For measurement of the finger movement, a bending sensor or indentation can thus be used as a command to control a device.63
rotary encoder can be used. Also, the muscle hardness change results in an alteration of the
natural frequency of the muscle. This change in natural frequency,
5.3 Breath Switch measured while providing oscillation with a vibrating element like a
Though they lack intuitiveness compared to other control piezoelectric material, can be regarded as a signal of muscle
methods, breath switches such as an expiration switch or a sip-and- activation.94
puff switch are also reliable means of controlling the assistive hand
exoskeleton.59,64 This method is especially useful for patients who 5.6 Mechanomyography (MMG)
have limited ability to control the device with their body motion or Mechanomyography (MMG) is a recording of the oscillations
activation of their skeletal muscles. which reflect the mechanical activities of the contracting muscle
caused by lateral dimensional changes of the active muscle fibers.95
5.4 Surface Electromyography (sEMG) Because the MMG signal reflects the number of recruited motor
Electromyography (EMG) is a technique for evaluation and units and their firing rates, it can be used to estimate the force
recording of the electrical activity produced by skeletal muscles.81 exerted by the skeletal muscles.96
In particular, surface electromyography (sEMG) is a noninvasive The use of MMG has some advantages over EMG. The
way to indirectly estimate the muscle activation level. The use of an placement of the MMG sensor does not need to be precisely
EMG signal as a command for control of an exoskeleton also has selected.97 Also, MMG is not influenced by changes in the skin
the advantage of eliminating the time delay generated when the impedance caused by sweat, because it is a mechanical signal.98
exoskeleton reacts to the human intention. This interface at a higher However, the non-stationary characteristics99 and nonlinearity100
level of the human neurological system makes it possible to make it difficult to use simple models for estimation of muscle
overcome the electro-chemical-mechanical delay which inherently force from MMG signals. Rather than adopt regression models,
exists in the musculoskeletal system.82 The time delay is the time ANN was used to estimate the muscle force.101
between the activation of the neural system and the actual onset of
movement of the muscles and the corresponding joints. When the 5.7 Photoplethysmography at Fingernail
EMG signal is used as a command input for device control, the The change of fingernail color that occurs when a human exerts
controller can acquire the neural activation information and process a gripping force can be used as a fingertip contact force sensor.102
it during the time interval. The collected EMG signal is processed When the fingertip contact force increases, the blood flow at the
for estimation of the user’s intention. The intention estimation, fingertip is altered. This alteration of the hemodynamic state results
resulting in an estimated joint torque or muscle force, is performed in modification of the fingernail color pattern. The color pattern
using a suitable model to represent the behavior of the muscle change is characteristically nonuniform along the length of the
according to the EMG signal. Studies have shown that the torque fingernail. These fingernail color patterns can be acquired by
developed by the related muscles can be estimated from the EMG photodetectors receiving the light from arrays of micro LEDs
signal.83-85 reflected from the fingernail.
Specifically, the sEMG signal from the forearm muscles has
been used for grip force estimation. Linear or nonlinear regression 5.8 Fingerpad Deformation
models can be used to estimate the grip force.86-88 Despite the When a fingertip is in contact with an object, the exerted
simplicity of these regression models, they can estimate the grip gripping force causes deformation of the fingertip skin. This
force well. An artificial neural network (ANN) can also be used for deformation can also be used as a contact force sensor.103 The
the estimation.89 The ANN assumes the muscle models as a black sensor is designed to be mounted on a fingernail without disturbing
box. This approach is useful because not all of the muscles related the haptic sensation at the fingertip. The width of the fingertip is
to the pinch force are located close to the surface, making the monitored using a strain gauge sensor.
 820 / MAY 2012 INTERNATIONAL JOURNAL OF PRECISION ENGINEERING AND MANUFACTURING Vol. 13, No. 5

5.9 Pressure Pattern (Force Myography) Review of Biomedical Engineering, Vol. 6, pp. 497-525, 2004.
A cuff with arrays of pressure sensors surrounding the forearm
2. Taub, E., Miller, N., Novack, T., Cook, E., Fleming, W.,
can be used to register the distributed mechanical force caused by the
Nepomuceno, C., Connell, J. and Crago, J., “Technique to
activation of the muscles. The pressures on the sensors are generated
improve chronic motor deficit after stroke,” Archives of
by volumetric changes in the underlying musculotendinous complex.
Physical Medicine and Rehabilitation, Vol. 74, No. 4, pp. 347-
From this force myography (FMG), individual finger movements can
354, 1993.
be encoded at the forearm in form of images for the control of robotic
and virtual hands for amputees.104-108 Grip force can also be estimated 3. Mark, V. W. and Taub, E., “Constraint-induced movement
from the summed and rectified FMG signals of the forearm.109 therapy for chronic stroke hemiparesis and other disabilities,”
Restorative Neurology and Neuroscience, Vol. 22, No. 3-5, pp.
317-336, 2004.
6. Conclusions
4. Patton, J. L. and Mussa-Ivaldi, F. A., “Robot-assisted adaptive
training: custom force fields for teaching movement patterns,”
With the advent of an aging society all over the world, there
IEEE Transactions on Biomedical Engineering, Vol. 51, No. 4,
will be increased demand for the practical application of assistance
pp. 636-646, 2004.
and rehabilitation technologies. Among the various possible body
parts, the hand may be the last endeavor for researchers because of 5. Heller, A., Wade, D. T., Wood, V. A., Sunderland, A., Hewer,
the many degrees of freedom and the number of tactile sensors in R. L. and Ward, E., “Arm function after stroke: measurement
the relatively small size of the part. and recovery over the first three months,” Journal of
Several research studies on hand exoskeletons have been Neurology, Neurosurgery, and Psychiatry, Vol. 50, No. 6, pp.
introduced in this paper. A summary of the fundamental 714-719, 1987.
technologies and challenges in the current research has also been
presented. It is promising that there are already some 6. Wade, D. T., Langton-Hewer, R., Wood, V. A., Skilbeck, C. E.
commercialized hand exoskeletons for rehabilitation applications. and Ismail, H. M., “The hemiplegic arm after stroke:
However, the development of assistive hand exoskeletons still has measurement and recovery,” Journal of Neurology,
many challenges to be overcome for practical usage. Neurosurgery, and Psychiatry, Vol. 46, No. 6, pp. 521-524, 1983.
It can be seen from our survey that most of the advanced work in 7. Sunderland, A., Tinson, D., Bradley, L. and Hewer, R. L.,
this field has been done in recent decades and many of the outcomes “Arm function after stroke. An evaluation of grip strength as a
have been demonstrated in a laboratory setting and in wired measure of recovery and a prognostic indicator,” Journal of
environments. Because not all of the technical components are well Neurology, Neurosurgery, and Psychiatry, Vol. 52, No. 11, pp.
developed enough or packaged for use in daily life and in outdoors 1267-1272, 1989.
applications, a considerable amount of cooperative work and use of
resources from medical technology, biomechanics, engineering, and 8. Nakayama, H., Jorgensen, H., Raaschou, H. and Olsen, T.,
product development are required. For outdoor use in particular, “Recovery of upper extremity function in stroke patients: the
power source technologies and reliable wireless technologies must be Copenhagen Stroke Study,” Archives of Physical Medicine
resolved. In fact, ensuring the portability of the hand exoskeleton and Rehabilitation, Vol. 75, No. 4, pp. 394-398, 1994.
system is possibly the most challenging part of the development. 9. Kapandji, I. A., “The physiology of the joints: annotated
This paper is intended to increase the focus on the hand diagrams of the mechanics of the human joints,” Churchill
rehabilitation and assistance device as an independent product and Livingstone, 1987.
provide the list of challenges in this area to enhance such small and
powerful devices. 10. Moran, C. A., “Anatomy of the Hand,” Physical Therapy, Vol.
69, No. 12, pp. 1007-1013, 1989.

11. Berme, N., Paul, J. P. and Purves, W. K., “A biomechanical

ACKNOWLEDGEMENT analysis of the metacarpo-phalangeal joint,” Journal of
Biomechanics, Vol. 10, No. 7, pp. 409-412, 1977.
This research was supported by the Happy tech. program through
the National Research Foundation of Korea (NRF) funded by the 12. Lluch, A., “Intrinsic causes of stiffness of the interphalangeal
Ministry of Education, Science and Technology (No. 2011-0020937). joints, in: Copeland, S. A., Gschwend, N., Landi, A. and Saffar,
P. (Eds.), Joint Stiffness of the Upper Limb,” Taylor & Francis,
pp. 259-264, 1997.
REFERENCES
13. Kapandji, I. A., “The Physiology of the Joints - Volume I:
Upper Limb, 5th ed.,” Churchill Livingstone, 1982.
1. Reinkensmeyer, D. J., Emken, J. L. and Cramer, S. C.,
“Robotics, motor learning, and neurologic recovery,” Annual 14. Hollister, A. and Giurintano, D., “Thumb movements, motions,
INTERNATIONAL JOURNAL OF PRECISION ENGINEERING AND MANUFACTURING Vol. 13, No. 5 MAY 2012 / 821

and moments,” Journal of Hand Therapy, Vol. 8, No. 2, pp. finger hand exoskeleton for VR grasping simulation,” Proc. of
106-114, 1995. the Eurohaptics, pp. 80-93, 2003.

15. Imaeda, T., An, K. and Cooney, W., “Functional anatomy and 28. Nakagawara, S., Kajimoto, H., Kawakami, N., Tachi, S. and
biomechanics of the thumb,” Hand Clinics, Vol. 8, No. 1, pp. Kawabuchi, I., “An Encounter-Type Multi-Fingered Master
9-15, 1992. Hand Using Circuitous Joints,” Proc. of the IEEE International
Conference on Robotics and Automation, pp. 2667-2672, 2005.
16. Barr, A. and Bear-Lehman, J., “Biomechanics of the wrist and
hand, in: Nordin, M. and Frankel, V. H. (Eds.), Basic 29. Wang, J., Li, J., Zhang, Y. and Wang, S., “Design of an
Biomechanics of the Musculoskeletal System, 3rd ed.,” exoskeleton for index finger rehabilitation,” Proc. of the
Lippincott Williams & Wilkins, pp. 358-387, 2001. Annual International Conference of the IEEE Engineering in
Medicine and Biology Society, pp. 5957-5960, 2009.
17. Taylor, C. L. and Schwarz, R. J., “The anatomy and
mechanics of the human hand,” Artificial Limbs, Vol. 2, No. 2, 30. In, H. K. and Cho, K. J., “Compact Hand Exoskeleton Robot
pp. 22-35, 1955. for the Disabled,” Proc. of the International Conference on
Ubiquitous Robots and Ambient Intelligence, 2009.
18. Elliot, D. and McGrouther, D. A., “The excursions of the long
extensor tendons of the hand,” The Journal of Hand Surgery: 31. Brokaw, E. B., Black, I., Holley, R. J. and Lum, P. S., “Hand
British & European Volume, Vol. 11, No. 1, pp. 77-80, 1986. Spring Operated Movement Enhancer (HandSOME): A
Portable, Passive Hand Exoskeleton for Stroke
19. Armstrong, T. J. and Chaffin, D. B., “An investigation of the Rehabilitation,” IEEE Transactions on Neural Systems and
relationship between displacements of the finger and wrist Rehabilitation Engineering, Vol. 19, No. 4, pp. 391-399, 2011.
joints and the extrinsic finger flexor tendons,” Journal of
Biomechanics, Vol. 11, No. 3, pp. 119-128, 1978. 32. Otto Bock HealthCare, “WaveFlex Hand CPM Device,”
http://www.ottobock.ca/cps/rde/xchg/ob_us_en/hs.xsl/15712.h
20. Kuczynski, K., “The proximal interphalangeal joint: anatomy tml
and causes of stiffness in the fingers,” Journal of Bone and
Joint Surgery-British Volume, Vol. 50, No. 3, pp. 656-663, 33. Patterson Medical, “Kinetec Maestra Portable Hand CPM,”
1968. http://www.pattersonmedical.com/app.aspx?cmd=get_product
&id=74161
21. Shrewsbury, M. and Johnson, R., “Ligaments of the distal
interphalangeal joint and the mallet position,” The Journal of 34. Mulas, M., Folgheraiter, M. and Gini, G., “An EMG-
Hand Surgery, Vol. 5, No. 3, pp. 214-216, 1980. controlled exoskeleton for hand rehabilitation,” Proc. of the
9th International Conference on Rehabilitation Robotics, pp.
22. Worsnopp, T. T., Peshkin, M. A., Colgate, J. E. and Kamper, D. 371-374, 2005.
G., “An Actuated Finger Exoskeleton for Hand Rehabilitation
35. Tong, K. Y., Ho, S. K., Pang, P. M. K., Hu, X. L., Tam, W. K.,
Following Stroke,” Proc. of the IEEE International
Fung, K. L., Wei, X. J., Chen, P. N. and Chen, M., “An
Conference on Rehabilitation Robotics, pp. 896-901, 2007.
intention driven hand functions task training robotic system,”
23. Fontana, M., Dettori, A., Salsedo, F. and Bergamasco, M., Proc. of the Annual International Conference of the IEEE
“Mechanical design of a novel Hand Exoskeleton for accurate Engineering in Medicine and Biology Society, pp. 3406-3409,
force displaying,” Proc. of the IEEE International Conference 2010.
on Robotics and Automation, pp. 1704-1709, 2009.
36. Iqbal, J., Tsagarakis, N. G., Fiorilla, A. E. and Caldwell, D. G.,
24. Wege, A. and Hommel, G., “Development and control of a “A portable rehabilitation device for the Hand,” Proc. of the
hand exoskeleton for rehabilitation of hand injuries,” Proc. of Annual International Conference of the IEEE Engineering in
the IEEE/RSJ International Conference on Intelligent Robots Medicine and Biology Society, pp. 3694-3697, 2010.
and Systems, pp. 3046-3051, 2005.
37. Schabowsky, C., Godfrey, S., Holley, R. and Lum, P.,
25. In, H. K., Cho, K.-J., Kim, K. R. and Lee, B. S., “Jointless “Development and pilot testing of HEXORR: Hand
structure and under-actuation mechanism for compact hand EXOskeleton Rehabilitation Robot,” Journal of
exoskeleton,” Proc. of the IEEE International Conference on NeuroEngineering and Rehabilitation, Vol. 7, No. 1, p. 36, 2010.
Rehabilitation Robotics, pp. 1-6, 2011.
38. Chiri, A., Giovacchini, F., Vitiello, N., Cattin, E., Roccella, S.,
26. Kadowaki, Y., Noritsugu, T., Takaiwa, M., Sasaki, D. and Vecchi, F. and Carrozza, M. C., “HANDEXOS: Towards an
Kato, M., “Development of Soft Power-Assist Glove and exoskeleton device for the rehabilitation of the hand,” Proc. of
Control Based on Human Intent,” Journal of Robotics and the IEEE/RSJ International Conference on Intelligent Robots
Mechatronics, Vol. 23, No. 2, pp. 281-291, 2011. and Systems, pp. 1106-1111, 2009.

27. Stergiopoulos, P., Fuchs, P. and Laurgeau, C., “Design of a 2- 39. Chiri, A., Vitiello, N., Giovacchini, F., Roccella, S., Vecchi, F.
 822 / MAY 2012 INTERNATIONAL JOURNAL OF PRECISION ENGINEERING AND MANUFACTURING Vol. 13, No. 5

and Carrozza, M. C., “Mechatronic Design and 50. Rotella, M. F., Reuther, K. E., Hofmann, C. L., Hage, E. B.
Characterization of the Index Finger Module of a Hand and BuSha, B. F., “An orthotic hand-assistive exoskeleton for
Exoskeleton for Post-stroke Rehabilitation,” IEEE/ASME actuated pinch and grasp,” Proc. of the IEEE Annual
Transactions on Mechatronics, Vol. PP, No. 99, pp. 1-11, 2011. Northeast Bioengineering Conference, pp. 1-2, 2009.

40. Wege, A. and Zimmermann, A., “Electromyography sensor 51. Baker, M. D., McDonough, M. K., McMullin, E. M., Swift, M.
based control for a hand exoskeleton,” Proc. of the IEEE and BuSha, B. F., “Orthotic Hand-Assistive Exoskeleton,”
International Conference on Robotics and Biomimetics, pp. Proc. of the IEEE 37th Annual Northeast Bioengineering
1470-1475, 2007. Conference, pp. 1-2, 2011.

41. Ueki, S., Kawasaki, H., Ito, S., Nishimoto, Y., Abe, M., Aoki, 52. Hasegawa, Y., Mikami, Y., Watanabe, K. and Sankai, Y.,
T., Ishigure, Y., Ojika, T. and Mouri, T., “Development of a “Five-fingered assistive hand with mechanical compliance of
Hand-Assist Robot With Multi-Degrees-of-Freedom for human finger,” Proc. of the IEEE International Conference on
Rehabilitation Therapy,” IEEE/ASME Transactions on Robotics and Automation, pp. 718-724, 2008.
Mechatronics, Vol. 17, No. 1, pp. 136-146, 2012.
53. Hasegawa, Y., Tokita, J., Kamibayashi, K. and Sankai, Y.,
42. Li, J., Zheng, R., Zhang, Y. and Yao, J., “iHandRehab: An “Evaluation of fingertip force accuracy in different support
interactive hand exoskeleton for active and passive conditions of exoskeleton,” Proc. of the IEEE International
rehabilitation,” Proc. of the IEEE International Conference on Conference on Robotics and Automation, pp. 680-685, 2011.
Rehabilitation Robotics, pp. 1-6, 2011.
54. Shields, B. L., Main, J. A., Peterson, S. W. and Strauss, A. M.,
43. Sarakoglou, I., Tsagarakis, N. G. and Caldwell, D. G., “An anthropomorphic hand exoskeleton to prevent astronaut
“Occupational and physical therapy using a hand exoskeleton hand fatigue during extravehicular activities,” Proc. of the
based exerciser,” Proc. of the IEEE/RSJ International IEEE Transactions on Systems, Man and Cybernetics, Part A:
Conference on Intelligent Robots and Systems, Vol. 3, pp. Systems and Humans, Vol. 27, No. 5, pp. 668-673, 1997.
2973-2978, 2004.
55. Yamada, Y., Morizono, T., Sato, S., Shimohira, T., Umetani, Y.,
44. Jones, C. L., Wang, F., Osswald, C., Kang, X., Sarkar, N. and Yoshida, T. and Aoki, S., “Proposal of a SkilMate finger for
Kamper, D. G., “Control and kinematic performance analysis EVA gloves,” Proc. of the IEEE International Conference on
of an Actuated Finger Exoskeleton for hand rehabilitation Robotics and Automation, Vol. 2, pp. 1406-1412, 2001.
following stroke,” Proc. of the 3rd IEEE RAS and EMBS
International Conference on Biomedical Robotics and 56. Benjuya, N. and Kenney, S. B., “Myoelectric Hand Orthosis,”
Biomechatronics, pp. 282-287, 2010. Journal of Prosthetics and Orthotics, Vol. 2, No. 2, pp. 149-
154, 1990.
45. Ren, Y., Park, H.-S. and Zhang, L.-Q., “Developing a whole-
arm exoskeleton robot with hand opening and closing 57. DiCicco, M., Lucas, L. and Matsuoka, Y., “Comparison of
mechanism for upper limb stroke rehabilitation,” Proc. of the control strategies for an EMG controlled orthotic exoskeleton
IEEE International Conference on Rehabilitation Robotics, pp. for the hand,” Proc. of the IEEE International Conference on
761-765, 2009. Robotics and Automation, Vol. 2, pp. 1622-1627, 2004.

46. Kinetic Muscles Inc., “Hand Physical Therapy with The Hand 58. Lucas, L., DiCicco, M. and Matsuoka, Y., “An EMG-
MentorTM,” http://www.kineticmuscles.com/hand-physical- controlled hand exoskeleton for natural pinching,” Journal of
therapy-hand-mentor.html Robotics and Mechatronics, Vol. 16, No. 5, pp. 482-488, 2004.

47. Takahashi, C. D., Der-Yeghiaian, L., Le, V., Motiwala, R. R. 59. Sasaki, D., Noritsugu, T., Takaiwa, M. and Yamamoto, H.,
and Cramer, S. C., “Robot-based hand motor therapy after “Wearable power assist device for hand grasping using
stroke,” Brain, Vol. 131, No. 2, pp. 425-437, 2008. pneumatic artificial rubber muscle,” Proc. of the IEEE
International Workshop on Robot and Human Interactive
48. Wu, J., Huang, J., Wang, Y. and Xing, K., “A Wearable
Communication, pp. 655-660, 2004.
Rehabilitation Robotic Hand Driven by PM-TS Actuators, in:
Liu, H., Ding, H., Xiong, Z. and Zhu, X. (Eds.), Intelligent 60. Tadano, K., Akai, M., Kadota, K. and Kawashima, K.,
Robotics and Applications,” Springer, Vol. 6425, pp. 440-450, “Development of grip amplified glove using bi-articular
2010. mechanism with pneumatic artificial rubber muscle,” Proc. of
the IEEE International Conference on Robotics and
49. Martinez, L. A., Olaloye, O. O., Talarico, M. V., Shah, S. M.,
Automation, pp. 2363-2368, 2010.
Arends, R. J. and BuSha, B. F., “A power-assisted exoskeleton
optimized for pinching and grasping motions,” Proc. of the 61. Takagi, M., Iwata, K., Takahashi, Y., Yamamoto, S. I.,
IEEE Annual Northeast Bioengineering Conference, pp. 1-2, Koyama, H. and Komeda, T., “Development of a grip aid
2010. system using air cylinders,” Proc. of the IEEE International
INTERNATIONAL JOURNAL OF PRECISION ENGINEERING AND MANUFACTURING Vol. 13, No. 5 MAY 2012 / 823

Conference on Robotics and Automation, pp. 2312-2317, 76. Abolfathi, P. P., “Development of an Instrumented and
2009. Powered Exoskeleton for the Rehabilitation of the Hand,”
Ph.D. Thesis, School of Aerospace, Mechanical and
62. Toya, K., Miyagawa, T. and Kubota, Y., “Power-Assist Glove
Mechatronic Engineering, University of Sydney, 2007.
Operated by Predicting the Grasping Mode,” Journal of
System Design and Dynamics, Vol. 5, No. 1, pp. 94-108, 2011. 77. Duncheon, C., “Robots will be of service with muscles, not
motors,” Industrial Robot: An International Journal, Vol. 32,
63. Moromugi, S., Koujina, Y., Ariki, S., Okamoto, A., Tanaka, T.,
No. 6, pp. 452-455, 2005.
Feng, M. Q. and Ishimatsu, T., “Muscle stiffness sensor to
control an assistance device for the disabled,” Artificial Life 78. Herr, H. and Kornbluh, R., “New horizons for orthotic and
and Robotics, Vol. 8, No. 1, pp. 42-45, 2004. prosthetic technology: artificial muscle for ambulation,” Proc.
of SPIE, Vol. 5385, pp. 1-9, 2004.
64. Makaran, J. E., Dittmer, D. K., Buchal, R. O. and MacArthur,
D. E., “The SMART Wrist-Hand Orthosis (WHO) for 79. Otsuka, K. and Wayman, C. M., “Shape memory materials,”
Quadriplegic Patients,” Journal of Prosthetics and Orthotics, Cambridge University Press, 1999.
Vol. 5, No. 3, pp. 73-76, 1993.
80. Mavroidis, C., Pfeiffer, C. and Mosley, M., “Conventional
65. Vas, P., “Sensorless vector and direct torque control,” Oxford Actuators, Shape Memory Alloys and Electrorheological
University Press, 1998. Fluids, in: Bar-Cohen, Y. (Ed.), Invited Chapter in Automation,
Miniature Robotics and Sensors for Non-Destructive Testing
66. Sul, S. K., “Control of electric machine drive systems,”
and Evaluation,” The American Society for Nondestructive
Wiley-IEEE Press, 2011.
Testing, pp. 189-214, 2000.
67. Kim, S. H., “DC, AC, BLDC motor control,” Bogdoo, 2010.
81. Kamen, G., “Electromyographic Kinesiology, in: Robertson, D.
68. Gopura, R. A. R. C. and Kiguchi, K., “Mechanical designs of G. E. (Ed.), Research Methods in Biomechanics,” Human
active upper-limb exoskeleton robots: State-of-the-art and Kinetics, 2004.
design difficulties,” Proc. of the IEEE International
82. Rosen, J., Brand, M., Fuchs, M. B. and Arcan, M., “A
Conference on Rehabilitation Robotics, pp. 178-187, 2009.
myosignal-based powered exoskeleton system,” IEEE
69. Tondu, B. and Lopez, P., “Modeling and control of McKibben Transactions on Systems, Man and Cybernetics, Part A:
artificial muscle robot actuators,” IEEE Control Systems Systems and Humans, Vol. 31, No. 3, pp. 210-222, 2001.
Magazine, Vol. 20, No. 2, pp. 15-38, 2000.
83. Triolo, R. J. and Moskowitz, G. D., “The theoretical
70. Noritsugu, T., Takaiwa, M. and Sasaki, D., “Development of development of a multichannel time-series myoprocessor for
power assist wear using pneumatic rubber artificial muscles,” simultaneous limb function detection and muscle force
Journal of Robotics and Mechatronics, Vol. 21, No. 5, pp. 607- estimation,” IEEE Transactions on Biomedical Engineering,
613, 2009. Vol. 36, No. 10, pp. 1004-1017, 1989.

71. Takashima, K., Noritsugu, T., Rossiter, J., Guo, S. and Mukai, 84. Clancy, E. A. and Hogan, N., “Relating agonist-antagonist
T., “Development of curved type pneumatic artificial rubber electromyograms to joint torque during isometric, quasi-isotonic,
muscle using shape-memory polymer,” Proc. of the SICE nonfatiguing contractions,” IEEE Transactions on Biomedical
Annual Conference, pp. 1691-1695, 2011. Engineering, Vol. 44, No. 10, pp. 1024-1028, 1997.

72. Bar-Cohen, Y., “EAP as artificial muscles: progress and 85. Nam, Y. S., Kim, S. N. and Baek, S.-R., “Calculation of Knee
challenges,” Proc. of the Smart Structures and Materials 2004: Joint Moment in Isometric and Isokinetic Knee Motion,” Int. J.
Electroactive Polymer Actuators and Devices (EAPAD), Vol. Precis. Eng. Manuf., Vol. 12, No. 5, pp. 921-924, 2011.
5385, pp. 10-16, 2004.
86. Duque, J., Masset, D. and Malchaire, J., “Evaluation of
73. Mirfakhrai, T., Madden, J. D. W. and Baughman, R. H., handgrip force from EMG measurements,” Applied
“Polymer artificial muscles,” Materials Today, Vol. 10, No. 4, Ergonomics, Vol. 26, No. 1, pp. 61-66, 1995.
pp. 30-38, 2007.
87. Hoozemans, M. J. M. and van Dieën, J. H., “Prediction of
74. Bar-Cohen, Y., “Electro-active polymers: current capabilities handgrip forces using surface EMG of forearm muscles,”
and challenges,” Proc. of the SPIE, the International Society Journal of Electromyography and Kinesiology, Vol. 15, No. 4,
for Optical Engineering, Vol. 4695, pp. 1-7, 2002. pp. 358-366, 2005.

75. Deole, U., Lumia, R., Shahinpoor, M. and Bermudez, M., 88. DiDomenico, A. and Nussbaum, M. A., “Estimation of forces
“Design and test of IPMC artificial muscle microgripper,” exerted by the fingers using standardised surface
Journal of Micro-Nano Mechatronics, Vol. 4, No. 3, pp. 95- electromyography from the forearm,” Ergonomics, Vol. 51,
102, 2008. No. 6, pp. 858-871, 2008.
 824 / MAY 2012 INTERNATIONAL JOURNAL OF PRECISION ENGINEERING AND MANUFACTURING Vol. 13, No. 5

89. Choi, C., Kwon, S., Park, W., Lee, H. and Kim, J., “Real-time “Upper trapezius muscle mechanomyographic and
pinch force estimation by surface electromyography using an electromyographic activity in humans during low force
artificial neural network,” Medical Engineering and Physics, fatiguing and non-fatiguing contractions,” European Journal
Vol. 32, No. 5, pp. 429-436, 2010. of Applied Physiology, Vol. 87, No. 4, pp. 327-336, 2002.

90. Yu, H. L., Chase, R. A. and Strauch, B., “Atlas of hand 101. Youn, W. and Kim, J., “Feasibility of using an artificial neural
anatomy and clinical implications,” Mosby Inc., 2004. network model to estimate the elbow flexion force from
mechanomyography,” Journal of Neuroscience Methods, Vol.
91. De Luca, C. J. and Merletti, R., “Surface myoelectric signal
194, No. 2, pp. 386-393, 2011.
cross-talk among muscles of the leg,” Electroencephalography
and Clinical Neurophysiology, Vol. 69, No. 6, pp. 568-575, 102. Mascaro, S. A. and Asada, H. H., “Photoplethysmograph
1988. fingernail sensors for measuring finger forces without haptic
obstruction,” IEEE Transactions on Robotics and Automation,
92. Martin, B. J., Armstrong, T. J., Foulke, J. A., Natarajan, S.,
Vol. 17, No. 5, pp. 698-708, 2001.
Klinenberg, E., Serina, E. and Rempel, D., “Keyboard
Reaction Force and Finger Flexor Electromyograms during 103. Nakatani, M., Shiojima, K., Kinoshita, S., Kawasoe, T.,
Computer Keyboard Work,” Human Factors: The Journal of Koketsu, K. and Wada, J., “Wearable contact force sensor
the Human Factors and Ergonomics Society, Vol. 38, No. 4, system based on fingerpad deformation,” Proc. of the IEEE
pp. 654-664, 1996. World Haptics Conference, pp. 323-328, 2011.

93. Lukowicz, P., Hanser, F., Szubski, C. and Schobersberger, W., 104. Abboudi, R. L., Glass, C. A., Newby, N. A., Flint, J. A. and
“Detecting and Interpreting Muscle Activity with Wearable Craelius, W., “A biomimetic controller for a multifinger
Force Sensors,” Pervasive Computing, Vol. 3968, pp. 101-116, prosthesis,” IEEE Transactions on Rehabilitation Engineering,
2006. Vol. 7, No. 2, pp. 121-129, 1999.

94. Kasuya, M., Seki, M., Kawamura, K. and Fujie, M. G., 105. Curcie, D. J., Flint, J. A. and Craelius, W., “Biomimetic finger
“Subtle grip force estimation from EMG and muscle stiffness control by filtering of distributed forelimb pressures,” IEEE
— Relationship between muscle character frequency and grip Transactions on Neural Systems and Rehabilitation
force,” Proc. of the Annual International Conference of the Engineering, Vol. 9, No. 1, pp. 69-75, 2001.
IEEE Engineering in Medicine and Biology Society, pp. 4116-
106. Craelius, W., “The Bionic Man: Restoring Mobility,” Science,
4119, 2011.
Vol. 295, No. 5557, pp. 1018-1021, 2002.
95. Barry, D. T. and Cole, N. M., “Muscle sounds are emitted at the
107. Kuttuva, M., Burdea, G., Flint, J. and Craelius, W.,
resonant frequencies of skeletal muscle,” IEEE Transactions on
“Manipulation Practice for Upper-Limb Amputees Using
Biomedical Engineering, Vol. 37, No. 5, pp. 525-531, 1990.
Virtual Reality,” Presence: Teleoperators and Virtual
96. Youn, W. and Kim, J., “Estimation of elbow flexion force Environments, Vol. 14, No. 2, pp. 175-182, 2005.
during isometric muscle contraction from mechanomyography
108. Phillips, S. L. and Craelius, W., “Residual kinetic imaging: a
and electromyography,” Medical and Biological Engineering
versatile interface for prosthetic control,” Robotica, Vol. 23,
and Computing, Vol. 48, No. 11, pp. 1149-1157, 2010.
No. 3, pp. 277-282, 2005.
97. Silva, J., Heim, W. and Chau, T., “A Self-Contained, 109. Wininger, M., Kim, N. and Craelius, W., “Pressure signature
Mechanomyography-Driven Externally Powered Prosthesis,” of the forearm as a predictor of grip force,” Journal of
Archives of Physical Medicine and Rehabilitation, Vol. 86, No. Rehabilitation Research and Development, Vol. 45, No. 6, pp.
10, pp. 2066-2070, 2005. 883-892, 2008.
98. Xie, H.-B., Zheng, Y.-P. and Guo, J.-Y., “Classification of the
mechanomyogram signal using a wavelet packet transform
and singular value decomposition for multifunction prosthesis
control,” Physiological Measurement, Vol. 30, No. 5, pp. 441-
457, 2009.

99. Akataki, K., Mita, K., Watakabe, M. and Itoh, K.,

“Mechanomyogram and force relationship during voluntary
isometric ramp contractions of the biceps brachii muscle,”
European Journal of Applied Physiology, Vol. 84, No. 1, pp.
19-25, 2001.

100. Madeleine, P., Farina, D., Merletti, R. and Arendt-Nielsen, L.,