Content uploaded by Adnan Rashid
Author content
All content in this area was uploaded by Adnan Rashid on Feb 12, 2018
Content may be subject to copyright.
Content uploaded by Adnan Rashid
Author content
All content in this area was uploaded by Adnan Rashid on Feb 06, 2018
Content may be subject to copyright.
Wearable Technologies for Hand Joints Monitoring for
Rehabilitation: A Survey
Adnan Rashid∗, Osman Hasan
School of Electrical Engineering and Computer Science (SEECS)
National University of Sciences and Technology (NUST), Islamabad, Pakistan
Abstract
Hand deformities often become a major obstacle in conducting everyday tasks
for many people around the globe. Rehabilitation procedures are widely used
for strengthening the hand muscles, which in turn leads to the restoration of
functionality of the affected hand. This paper conducts a survey of various
wearable technologies that can be used to accurately quantify the rehabil-
itation progress in terms of fingers’ hand joint angles. Based on the data
acquisition methods, these technologies can be mainly divided into six cat-
egories: 1) Flex sensor based; 2) Accelerometer based; 3) Vision based; 4)
Hall-effect based; 5) Stretch sensor based; and 6) Magnetic sensor based. The
main focus of some of the discussed technologies has been on various other
domains, like gaming gloves, tele-manipulation etc., and thus their usage for
rehabilitation of hand joints could be quite interesting. This paper analyzes
the strengths and weaknesses of these wearable technologies along with some
examples of their implementations. Based on our survey results, we propose
a wearable glove for accurately measuring hand joint angles with enhanced
features for better diagnosis and rehabilitation.
Keywords: Flex Sensors, Magnetic Sensors, Accelerometers, Hall-effect,
Stretch Sensors, Hand Joints, Gloves, Dexterity.
∗Corresponding author
Email addresses: adnan.rashid@seecs.nust.edu.pk (Adnan Rashid),
osman.hasan@seecs.nust.edu.pk (Osman Hasan)
Preprint submitted to Microelectronics Journal January 20, 2018
1. Introduction
Many acquired hand deformities, such as Osteoarthritis, fractures due
to injuries, ruptured ligaments and dislocations, Rheumatoid Arthritis (RA)
and Carpal Tunnel Syndrome (CTS) [1], continue to affect the lives of many
humans around the globe. For example, it has been reported in [2] that up
to 50% of the people suffering from RA lose their jobs within the first 5 years
of diagnosis and the annual medical cost of an RA patient was estimated
at £3,600 in 1992 and it became about £4,000 per annum since 2011 [2].
Similarly, it is believed that 3.8% of the general population suffers from CTS
around the globe [3]. CTS is reported [3] to have been diagnosed in every
1 out of 5 persons who suffers from the pain and numbness in the hands,
which represents the severity and sensitivity of this problem. The above-
mentioned acquired hand deformities are generally cured by occupational
therapy prescribed by hand therapists. Treatment for hand deformities is
planned by doctors and clinicians after performing a complete check-up of
the patient using radiative techniques, such as X-rays, or manual techniques,
such as inspection of hand, health assessment questionnaires and examining a
range of motion of all hand joints. The prescribed treatment mainly consists
of physical therapy of hands to strengthen the muscles and thus recover the
lost functionality of joints. One of the major goals of the treatment is to
relieve pain and restore the functionality of hand for which hand therapists
use joint protection exercises and work routines, as illustrated in Figure 1.
Figure 1: Various Hand Movements [4]
Traditionally, physiotherapists use Goniometers [5] (for hand angles mea-
surement), hand strength dynamometers [6] (for hand grip force measure-
ment) and assessment questionnaires to quantify disease progression and to
monitor the rehabilitation process, i.e., the ability of the patient to perform
2
different tasks. Statistical analysis techniques are then applied based on
these outcomes to calculate patient’s hand functionality level. The outcomes
of this kind of manual procedure are easily influenced by the level of training
and experience of clinicians as the patients data is recorded in manual form.
Moreover, patients have to visit the clinic every time they want to check their
progress, which not only makes the whole process very time consuming but
also raises the burden on the healthcare costs.
The recent developments in wearable and Internet-of-Things (IoT) tech-
nologies can alleviate many of these issues by providing a more accurate,
reliable and automated solution for quantifying the rehabilitation progress.
These technologies can be broadly categorized into six types. (1) Flex Sen-
sor based technology, in which different types of resistive bend sensors are
embedded onto a stretchable glove and calibrated for accurate hand joint
measurement. (2) Accelerometer based technology, in which accelerometers
are similarly placed on a glove and calibrated for accurate readings. (3)
Vision based technology, in which a glove is specifically designed with dif-
ferent colours to employ gesture recognition algorithms with cameras. (4)
Hall-effect Sensor based technology, in which hall-effect sensors are used to
accurately measure the flexion/extension and abduction-abduction motion
of the fingers’ proximal joints. (5) Stretch Sensor based technology, in which
the sensor’s deformation, i.e., stretching and squeezing, provides the accurate
measurement of the hand and finger’s motion and joint measurement. (6)
Magnetic Sensor based technology, in which magnetic field sensors are used
to track the position and orientation of the hand. All these techniques can be
used to make a comprehensive wearable device, which can help in alleviating
the inaccuracies caused by the above-mentioned manual approach of rehabil-
itation quantification. Moreover, the ability of the patient to independently
measure and data log his/her rehabilitation progress using these automatic
methods tends to increase the quality of diagnosis and the treatment.
In this paper, we provide an extensive survey of all these technologies
along with some associated prototypes for hand joint monitoring. Our main
objective is to identify the advantages and drawbacks of these technologies
and the corresponding prototypes in order to recognize potential room for
improvement in research in the domain of accurate hand joint measurement
for rehabilitation quantification using the wearable technology. Moreover,
based on the analyses, we propose a wearable device to accurately measure
hand joint angles by utilizing the best possible features from all the mentioned
technologies. The proposed device includes conductive ink based sensors for
3
finger joints measurement along with an accelerometer for the measurement
of wrist joint angles. Also, the inclusion of a smartphone app for easy user
interface and automatic data entry can be used. We also recommend to use
this wearable device to measure the level of dexterity of a patient’s hand.
This can be done by performing dexterity tests while donning the wearable
and getting necessary readings [7].
2. Flex Sensor Based Technologies
Flex sensors are passive resistive devices [8], which are commonly used to
measure angle of deflections. Flex sensors are generally composed of carbon
resistive elements, which are present within a flexible substrate. A bend in
a flex sensor results in a change in carbon content in the substrate, which
leads to a proportional change in the resistance of the substrate. Due to this
characteristic, flex sensors are also commonly termed as analog resistors.
Figure 2 shows a gesture recognition glove with flex sensors embedded on its
fingers [9].
Figure 2: An Example of a Flex Sensor Based Glove [9]
Flex sensors can be manufactured based on the conductive ink or the
fiber-optic technologies. Conductive ink sensors are fabricated by laying re-
sistive ink on a substrate. As the flex sensor is bent, the resistive material is
4
pulled apart and its resistance changes. In comparison to this, the fiber op-
tic sensor consists of a plastic fiber optic, a light source and a photosensitive
receiver. Light is sent from one end of the fiber optic cable and received at
the other. When the optical fiber is bent, light intensity at the receiving end
changes and thus the bending angle can be detected. Both conductive ink
and fiber-optic based flex sensors have less hysteresis in resistance, but con-
ductive ink based sensors are usually cheaper to manufacture than the fiber
optic ones. The conductive ink based flex sensors can bear slightly higher
temperature and humidity conditions while the fiber-optic based sensors have
high repeatability. Table 1 shows a comparison of these two commonly avail-
able conductive ink based flex sensors.
Table 1: Comparison of Flex Sensors [10, 11]
Characteristics Flex Point Spectra Symbol
Life Cycle >1 Million Bends >1 Million Bends
Temperature range -35◦C to +80◦C -35◦C to +80◦C
Flat Resistance 100 to 500K Ohms 25K Ohms
Resistance Tolerance/
Nonlinearity
NA 30 %
Bend Resistance Range 1.5K to 40K Ohms 45K to 125K Ohms
Power Rating NA 0.50 Watts Continuous
Hysteresis 7 % NA
Resolution <1 degree <1 degree
Operating Voltage 5V to 12V NA
2.1. Relevant Work
Flex sensor based technology [8] is the most widely used method in de-
signing wearables associated with hands. Cyberglove III [12] is a flex sensor
based glove used for gaming purposes and PC control. It has 18-22 sensors
embedded on it with a reasonable accuracy of <1 degree. Kumar et al. [13]
used a similar kind of glove, named DG5 VHand 2.0, for gesture recognition
used with gaming consoles. These gloves provide reasonable alternatives for
input devices for gaming consoles but are not suitable for hand rehabilita-
tion since their sensors are not placed for measuring the corresponding joint
movements. Recently, some researchers have explored the option of using
the flex sensor technology for measuring hand joint movements. For ex-
ample, Connolly et al. [14] proposed two flex sensor based gloves for hand
5
joints measurement, namely 5DT Data Glove and X-IST Data Glove. The
idea, presented in these two gloves, was extended by the researchers at the
Tyndall Institute in Ireland, for the development of a more sophisticated
wearable glove [15], which provides more accurate hand joint readings along
with other parameters of hand deformity quantification.
Gallo et al. [16] proposed a low cost portable user interface for 3D visu-
alization and manipulation of a medical image in a semi-immersive virtual
environment, which helps the radiologists in the understanding of the shape
and position of the anatomical structures. This 3D interface uses a Wiimote-
enhanced wireless data glove as an input and thus provides an exploration
of a 3D medical image in a virtual environment, such as rotation and move-
ment of 3D reconstructions of anatomical parts, and position control of 3D
cursor. The proposed interface used DG5 VHand 2.0 data glove [17], which
consists of one bend flex sensor on each of the finger and an accelerome-
ter for hand movement sensing and hand orientation deduction along the
3 main axes. Zimmerman et al. [18] proposed a hand to machine interface
device, which provides hand’s position and orientation information, and ges-
ture recognition. It mainly consists of a glove, which uses analog flex sensors
for measuring the finger bend, ultrasonics to measure the hand’s position
and orientation. Moreover, a small cable is used to establish a connection
between these sensors and the driving hardware. A custom sensor glove is
proposed in [19], which used passive-resistive flex sensors for the real-time
measurement of the finger flexion in the individuals having reduced range
of the motion of their hands and fingers. This glove is used to capture the
daily routine rehabilitation activities performed away from the clinical sites.
Saggio [20] proposed a novel array of flex sensors, which is integrated in
a sensory glove and can perform the goniometric semi-automated measure-
ments. These flex sensors are basically developed by coating resistive carbon
elements on a flexible thin plastic substrate. These are low cost with long
mechanical durability and can provide a good electrical stability over the
time.
Zhang et al. [21] developed a low cost viable rehabilitation system for
hand motion using the Augmented Reality (AR) technology. A self-designed
low cost data-glove is used for the interaction between the real hand and the
virtual environment and thus for the detection of flexion of the fingers and
to control the movements of the virtual hands. A novel pneumatic glove,
the PneuGlove [22], is used for hand rehabilitation after a stroke. It uses
Flexpoint sensors for measuring the joint kinematics and thus is used for
6
training of the grasp-and-release movements in a virtual reality environment.
A detailed account of the flex-based sensors and the gloves using these sensors
with their applications can be found in [8] and [23], respectively.
2.2. Discussion and Analysis
Flex sensors are very suitable for measuring hand gestures and a range
of movement of joints due to their ease of placement on a glove, their larger
life cycle, wide range of temperature of operation and finally, easy market
accessibility. They have a thin, flexible membrane that can be easily placed
on the glove over the knuckle of the joint under observation, providing a life
cycle of greater than 1 million complete bends. The ability to operate in
a wide range of temperature (-35◦C to +80◦C) makes them suitable for all
environments. Flex sensors are also available in different sizes, which makes
them quite a suitable choice for measuring different joints of hand.
Flex sensors also have some disadvantages, including the problem of re-
peatability and decrease in their accuracy over time. Also, the Flexpoint
sensors show a non-linear trend for smaller angles, which makes their cali-
bration quite difficult [8]. Bending a flex sensor with no protective coating
for a relatively longer period of time can result in a permanent bend in the
sensor that affects its base resistance, and requires a recalibration. More-
over, the flex sensors exhibit moderately slow response time due to their
physical deformation [24]. For instance, the resistive flex sensor has a typical
response time of 1 to 2 ms [8]. Another limitation of flex sensors is that they
can only measure bending angles of bodies with one degree of freedom while
accelerometers, vision based sensors and magnetic sensors can do the same
in more than one degree of freedom.
3. Accelerometer Based Technologies
Accelerometers are used to measure the orientation of an object [25].
They are used collectively with gyroscopes and magnetometers to form an
Inertial Measurement Unit (IMU), which provide quite accurate readings
of orientation. The most commonly used accelerometers are Micro-electro-
mechanical Systems (MEMS), which track the orientation based on the move-
ment of a small proof mass on a silicon surface, suspended by small beams.
The acceleration is measured based on the Newton’s second law of motion,
F=ma, in this setup as the beams act as springs. The second major type of
7
accelerometers is based on the piezoelectric technology, where the accelera-
tion changes in direct proportion to the applied force due to the piezoelectric
effect, which states that a charge of opposite polarity appears on opposite
sides of a certain crystal when crystals are compressed [26]. Table 2 shows a
comparison of both of these accelerometers.
Table 2: Piezoelectric Accelerometer Performance Compared with ADXL105 [14]
Property Piezoelectric MEMS (ADXL105)
Range Up to 2000 g 5 g
Sensitivity 100 pC/g 250 mV/g
Noise Density 0.02 mg 0.225 mg
Temperature Range 274◦C to 2508◦C 240◦C to 858◦C
Frequency Range 0.1 Hz to 4800 Hz 0 Hz to 10000 Hz
Resonance 16 kHz Around 7 kHz
3.1. Relevant Work
Just like the flex sensor based gloves, the main application of accelerome-
ter based hand gloves is in gaming and gesture recognition. KeyGlove [27, 28],
shown in Figure 3, Gest [29] and Acceleglove [30] use accelerometers to pro-
vide gesture recognition capabilities for PC control and gaming consoles. The
accuracy of these gloves is up to a few degrees, which makes them not too
suitable for precise measurement of hand joint angles, which is a require-
ment for rehabilitation. Hsiao et al. [28] proposed a data glove embedded
with 17, 9-axis IMUs. These IMUs contain a 3-axis accelerometer, a 3-axis
gyroscope and a 3-axis magnetometer to gauge acceleration, angular velocity,
and magnetic field, respectively. These sensors are placed on both the front
and back of the glove. The sensors are calibrated using an inertial motion
sensor to check the correctness of raw data. The acquired readings are vali-
dated by goniometers and servo motors and show an error of approximately
0.98 degrees. This glove was also used for clinical testing for dexterity results
and showed very clear and precise readings. However, it requires a flexible
printed circuit, which is an expensive technology for a consumer product.
The accelerometers have also been used for gesture recognition, which
is used in many applications, such as human computer interaction and sign
language translation. Bui et al. [32] developed a MEMS accelerometer based
8
Figure 3: An Example of an Accelerometer Based Glove [31]
glove for gesture recognition. It consists of six MEMS accelerometers, i.e., one
of them is placed at the back of the hand alongside the other five sensors at
the front of the hand, which considerably improves the process of recognition.
However, it can only perform gesture recognition in two dimensions. In order
to overcome this limitation, Kim et al. [33] developed a data glove KHU-1,
which consists of three tri-axis accelerometer sensors and thus can perform
a 3-D rule based hand motion tracking and gesture recognition. The signal
produced as a result of the process of sensing is transmitted to a computer
via a wireless channel using Bluetooth technology, which is used for further
processing and recognition process. Similarly, Zu et al. [34] presented three
models, which used the MEMS 3-axes accelerometers to recognize, seven
hand gestures including right, left, up, down, cross, tick and circle. The
sensed acceleration of the hand is transmitted via a Bluetooth channel to
the computer and gesture recognition is performed based on a comparison of
gesture code with the already stored template in the computer.
Hernandez-Reboller et al. [35] presented an interactive computer game
AcceleSpell, which helps in learning and practicing finger spelling. This game
is based on a decision tree based recognition algorithm and the AcceleGlove
glove. AcceleGlove utilizes six 3-axis accelerometers placed at the fingers and
back of the palm to provide the angular position of each axis upon a query
from the PC. OFlynn et al. [36] developed an Inertial Measurement Unit
(IMU) smart glove microsystem based on sensors, processors and wireless
technology used for human computer interaction. It consists of 16 9-axes
IMUs, where each one includes a 3-axis accelerometer, a 3-axis gyroscope
9
and a 3-axis magnetometer and provides a realtime measurement of a range
of hand joint movements including the measurement of flexion/extension,
adduction-abduction and complex hand movements. A detailed review of
the utilization of the accelerometer based sensors in wearable devices can be
found in [37]. Similarly, a review of wearable sensors and systems used in
rehabilitation can found in [38].
3.2. Discussion and Analysis
The accelerometer based technology for tracking hand joint movements,
explained above and the prototypes associated with it also has some advan-
tages and shortcomings. The advantages include less hardware requirement
and a better data rate as accelerometers give digital output and thus, do not
require analogue to digital conversion. Also, accelerometers are relatively
cheaper and have a longer lifespan. Moreover, accelerometers generally have
a faster response time compared to that of the flex sensors. For example, the
KC-2105 accelerometer has a response time of <100ns [39].
The disadvantages of this technology include the tricky placement of the
sensors on the glove. For assessing the movements of every hand joint an-
gle, accelerometers have to be placed in between each finger joint. This is
quite challenging due to the fixed shape and dimensions of the accelerome-
ters. Moreover, accelerometers provide their readings with reference to the
gravitational acceleration, i.e. g. This results in a lot of noise in the read-
ings and hence requires noise reduction algorithms, which in turn require
tedious initial calibration along with inaccuracies in readings in the presence
of magnetic devices.
4. Vision Based Technologies
Using imaging cameras for recognizing hand gestures, is a growing re-
search trend these days. Real-time hand tracking devices are used to capture
the freeform motion of the hand and the captured images are then used to
measure the hand joint angles and a range of movements using image pro-
cessing techniques as illustrated in Figure 4. Hand gesture recognition can be
primarily modeled by using 3D model based or appearance based techniques.
The 3D model method is primarily based on a 3D kinematic model of the
hand [40]. The required parameters for palm position and joint angles can
be obtained from this 3D information using volumetric models. The main
idea is to deduce hand parameters by comparing the possible 2D appearance
10
as projected by the 3D hand model and the input image from the camera.
The appearance-based techniques use images or videos as their input and
the required parameters are extracted using a template database instead of
using a spatial representation of the gesture [40]. Color body markers [41]
have also been used to track motion of the hand by using particle filtering
and multi scale color features [42].
Figure 4: An Example of a Vision Based Glove [43]
4.1. Relevant Work
Kapuscinski [44] used the skin colored section of the captured image to
intensity-normalize it with the desired hand region. This way the gesture
can be recognized using a Hidden Markov model. YCbCr color model was
used by Yu to differentiate skin colored pixels from the background of the
image [45]. Malima [46] used the red-green ratio of the image to detect the
skin colored portion. This way the hand’s center of gravity is determined,
which in turn allows us to find the location of the finger tips that is usually
the farthest point from the center of gravity point of the hand. A circle can
then be drawn around the center of gravity to determine the desired ges-
ture by counting the number of white pixels outside the circle. Jackin [47]
11
used a similar gesture recognition procedure, but instead of converting Red,
Blue and Green (RGB) to Hue, Saturation, and Value (HSV), as done by
Malima [46], they used RGB as input. This technique is reported to achieve
an accuracy of 100%. Koh [48] considered the shape and color of an image
for the identification of rough contour of the hand. Fang [49] employed the
Adaptive Boost algorithm to detect the hand from the input image. In ad-
dition to detecting a single hand, this algorithm can also detect overlapping
hands. Rekha et al. [50] detected the palm and finger structure by drawing
blobs and ridges and the recognition rate was found to be about 98% accurate.
Zabulis et al. [51] proposed a vision-based hand gesture recognition sys-
tem used for human computer interaction. It is based on a probabilistic
framework, which detects the image regions belonging to human hands effi-
ciently using multiple information clues. The process of real-time tracking
can handle multiple hands moving in different trajectories in front of the cam-
era. The authors [51] proved the efficiency and effectiveness of the proposed
approach using several experimental results. Cameiro et al. [52] developed
a Rehabilitation Gaming System (RGS), i.e., a vision based motion capture
system, which can be used for the rehabilitation of the stroke patient. Placidi
et al. [53] proposed a virtual glove, which unlike the traditional mechanical
gloves, uses the video cameras for capturing and tracking movements of the
hand. It uses a numerical hand model to calculate the physical and geo-
metrical parameters using some boundary constraints, i.e., joint angles and
dimensions. The proposed system is low cost and easy to use. Similarly, Ma
et al. [54] developed a five-fingered haptic glove, which is adaptable to any
size of the fingers and can provide accurate tracking the complex finger joint
motion. It uses least squares fitting of circles methods for the analysis of
the kinetic model of hand and fingers motion. It can be used for the reha-
bilitation of the hand as well as in virtual reality based systems. A detailed
account of the applications of the vision-based hand gestures can be found
in [55].
4.2. Discussion and Analysis
3D Model Based approach is computationally quite intensive and thus,
using it for real-time data acquisition, requires high performance computa-
tional resources, such as processing speeds and memory. Performance evalu-
ation for hand gesture recognition techniques can be done by evaluating the
12
percentage of error in recognition. Table 3 compares the accuracy of different
vision based hand gesture recognition techniques.
Table 3: Comparison of Vision Based Hand Gesture Recognition Techniques
Available Hand Gesture Recognition Tech-
niques
Accuracy
Hand gesture recognition by Hit-Mass Transform and
Hidden Markov Model [45]
98%
YCbCr color model with Artificial Neural Network
(ANN) [46]
97.4%
Detection of skin region by Red/Green ratio [47] 98%
Hand gesture recognition through Perceptual Color
Space [48]
100%
On-premise skin colored modeling method [49] 82.6%
Adaptive Boost Algorithm for hand detection [50] 98%
Hand Gesture recognition using PCBR and 2-D WPD
techniques [43]
91.3% static
86.3% dynamic
Two hand segmentation with Haar-Like feature and
adaptive skin color model [56]
89% to 98%
A few studies have also compared the accuracy of vision based tech-
niques with sensor based techniques. Table 4 shows the results of one such
experiment, conducted by Baatar et al. [57], involving 5 male and 5 female
participants.
Table 4: Experiment conducted to determine the accuracy of Vision based techniques
Sensor based hand
gesture recognition
Vision based hand
gesture recognition
Accuracy 84% 91%
User Preference 40% 60%
5. Hall-effect Sensor Based Technologies
Hall-effect sensors [58, 59] are used for the accurate measurement of the
flexion/extension and adduction-abduction motion of the proximal joint of
13
the fingers. These sensors are based on the phenomenon of the magnetic field,
which is characterized by its polarity and the flux density. When a magnetic
field is applied across a hall-effect sensor, its magnetic flux density starts
increasing. As this density crosses a pre-set threshold, the sensor detects it
and generates an electrical signal as an output voltage known as hall voltage.
The generation of this electrical signal based on the applied magnetic field
is known as the hall effect. Figure 5 depicts the Humanglove having 20 hall-
effect sensors, which is mainly used to measure the flexion/extension of the
fingers and thumbs as well as their adduction-abduction motion.
Figure 5: An Example of a Hall-effect Sensor Based Glove [60]
Based on the type of the output signal, these sensors are characterized
namely as analog and digital. In the analog sensor, the output signal is of
continuous nature and is directly proportional to the strength of the applied
magnetic field. The increase in the strength of the applied magnetic field
increases the corresponding output voltage until it saturates due to the lim-
itation applied on it by the power supply. Similarly, in the case of digital
sensors, it works as a switch, i.e., if the magnetic field crosses a pre-set value,
the output of the sensor switches from state “OFF” to “ON”. Moreover,
based on the utilization of the magnetic poles (north and south), the digital
hall-effect sensors are categorised as unipolar and bipolar sensors.
14
5.1. Relevant Work
The hall-effect sensor based wearable technology is widely used for track-
ing various motions of hands and fingers in many applications, such as
robotics, health-care and wearable computers. Humanglove [61] is a hall-
based sensor glove, which is used for measuring the flexion/extension and
abduction-abduction of the four fingers and thumb. It is embedded with 20
hall-effect sensor that are used for the above-mentioned measurement and
is available in 3 sizes. It is calibrated for the new user using the Graphical
Virtual Hand (GVH) software, which captures the actual movement of the
hand by its equivalent animated hand [62]. Hall-effect sensors are integrated
in the hand exoskeleton exerciser, which is used for the rehabilitation of pa-
tients who loose the muscular control of their hands [63]. Huang et al. [64]
proposed a wearable rehabilitation robotic hand that can be worn on the
forearm. Hall-effect sensor is embedded at the axis of the joints for the mea-
surement of the angles. A wearable artificial hand is proposed in [65], which
is used for prosthetics and humanoid robotics applications. In order to sense
and measure the angular movement and position of the hand joints, it uses
six hall-effect based sensors (SS495A, Honeywell, USA [66]). Due to their
small sizes and the contactless working principle, these sensors enable the
smooth working of the system by avoiding the frictional forces.
Hall-effect sensor grid is used on the fingernail to convert it into a touch-
pad, named as FingerPad, and a magnet at the thumb enables it to work as
a touch pad [67]. Similarly, Phillips et al. [68] proposed a transducer, which
is used for the monitoring of the patients undergoing the rehabilitation of
the flexor tendon of the hand fingers. It uses the hall-effect sensor to capture
the movement and stretching of the fingers. Chouhan et al. [69] presented
a glove for gesture recognition that can be used by the hearing and speech
impaired people. This system uses the hall-effect based sensor along with
other sensors to capture the hand and finger orientation and gestures.
Eilenberg et al. [70] presented an adaptive muscle-reflex controller for
anklefoot prostheses, in which a linear hall-effect sensor (Allegro A1395 [71])
is used for the estimation of the ankle joint angle, which ranges from -0.19
to +0.19 radians. Similarly, Arami et al. [72] proposed a hall-effect based
sensors system for the accurate measurement of the knee flexion-extension.
These sensors are integrated into a smart knee prosthesis, and are used to
simulate the actual patterns of walking.
15
5.2. Discussion and Analysis
Hall-effect sensors are low cost. Moreover, they are not effected by envi-
ronmental impurities due to their strong sealed packaging and thus can bear
severe conditions. The operating frequency for such sensors is up to 100kHz
and are thus very good for a high speed operation. These sensors can work in
a wider range of temperature and thus can measure a wider range of magnetic
fields. These sensors maintain their quality and are usable for an unlimited
period of time and thus can perform repeatable operations.
As described above, these sensors work on the phenomenon of magnetic
field, so there is always a possibility of the interference of this magnetic
field with the external magnetic field, which can change the resulted output.
This in turn may result in the degradation in performance by compromising
the accuracy of the sensed signal. Moreover, the response time for the hall-
effect sensors is slower compared to that of the accelerometers and it typically
ranges from 1 to 6µsec. For example, the VF360NT sensor exhibits a response
time of 1.5µsec [73].
6. Stretch Sensor Based Technologies
Stretch sensors [74] are used to measure stretch, bend, pressure and force
and are widely used for tracking hand movements in applications ranging
from soft robots, Virtual Reality (VR) gloves, biometric displacement read-
ing and other physical applications. These sensors are typically resistors
with resistance values depending on the sensor’s deformation, i.e., stretching
or squeezing. The deformation of the sensor is directly proportional to its
resistance, i.e., its stretching increases its resistance, whereas its squeezing
decreases its equivalent resistance [75]. In order to perform different stretch
sensor based operations, we need to adapt the same methodology, which is
used for the measurement of the resistance of a variable resistor. These sen-
sors are available in different sizes, sensitivities and elasticities based on their
intended applications. Figure 6 illustrates a soft stretchable bending sensor
placed at a hand’s finger.
Based on the process of their fabrication, the stretch sensors are broadly
characterized as of three types [76], namely fabric stretch [77], constructed
stretch [78] and the knit stretch [79] sensors. The fabric stretch sensors are
made up of stretchy fabrics, which are coated with a conducting polymer
named polypyrrole. Table 5 presents the specifications of some of the fabric
stretch sensors, which help in selecting an appropriate sensor based on its
16
Figure 6: A Soft Stretchable Bending Sensor Placed at a Hand’s Finger [74]
utility in a particular application. The constructed stretch sensors are devel-
Table 5: Specification of Some of the Fabric Stretch Sensor [77]
Batch Number Resistance (Range in Ohms) Size
RL-5-123-T-24 1100 to 1300 9”×11”
RL-5-109-GL 1500 to 2000 9”×11”
RL-5-129 8 ×105to 1 ×1069”×12”
RL-5-137 5 ×105to 1 ×10610”×11”
RP-3-125-1B 1100 to 1300 9”×17”
RP-3-125-2B 120 to 145 9” ×15”
oped using various techniques based on changing their conductive properties,
i.e., as a result of, knitting and stitching with resistive thread and a mixture
of conductive fiber with stretchy fabric glue. Similarly, the knit stretch sen-
sor is a 3×6 cm hand knitted rectangular conductive/resistive wool obtained
from a conductive thread, which can be embedded on any glove. If a circular
sewing machine is used, it will result into a circular knit stretch sensor [80].
17
6.1. Relevant Work
Stretch sensor based technology is widely used for measuring the stretch,
bend and force in soft robots and VR gloves, which are used for the re-
habilitation process. Toomey et al. [81] carried out study on the usage of
Dielectric Electroactive Polymer (DEAP) along with the stretch sensor in
human healthcare. The authors explored the integration of soft material and
the sensors for the real-time monitoring and mapping of the human body. A
fabric-based, flexible and stretchable tactile sensor is proposed in [82]. Next,
Bianchi et al. [83] presented a sensing glove for hand monitoring by integrat-
ing the kinaesthetic and stretchable tactile sensing gloves proposed in [82].
The stretchable tactile sensing glove consists of a knitted piezoresistive fab-
rics and is used to measure bending. Only five sensors, one placed on each
of the fingers enables the reconstruction of the 19 Degrees of Freedom (DoF)
hand models. A virtual reality-based system is proposed for the analysis of
accurate hand functionality in [84]. In the proposed system, the authors used
a glove for capturing the movement/motion of the hand, which was equipped
with the stretch sensors and has been used to trace the positions of the joint
exertion. Lee et al. [85] fabricated a stretchable strain sensor for the detec-
tion of tensile as well as compressive strains. This sensor is successfully used
in the human wrist and finger motions. It can also be used to measure the
pressure with high sensitivity.
Tognetti et al. [86] developed a prototype, which involves the knitted
and woven e-textile stretch sensors and is used to monitor the rehabilitation
progress of the knee joint. Similarly, Shimada et al. [87] used a stretch sensor
in the closed-loop system for the restoration of standing in paraplegia. Shen
et al. [74] presented a soft stretchable bending sensor and two sensor gloves.
Due to their low cost and customizable size, they are widely used in wearable
devices. A detailed account of the stretch sensor based technologies for the
rehabilitation purposes can be found in [88].
6.2. Discussion and Analysis
Due to the stretching ability, stretch sensors are customizable in size and
can fit any application. For example, in the case of non-stretchable data
gloves, the size of a glove is kept larger than the size of the hand so that
the sensors can fit into it. Thus, this gap between the hand and glove can
produce errors in the sensor’s output. Whereas, in the case of stretchable
glove, i.e., the glove having stretchable sensors, we can use the size of a glove
18
same as that of a hand and hence, can minimize the error caused by the
non-stretchable gloves.
One of the drawbacks of knit stretch sensor is that it follows the axis of the
knit structure and thus a free-form sensing pattern cannot be created. The
other drawback of the stretch sensor is that the sensitivity of these sensors
changes with the size of the sensor and hence is not constant. Moreover, these
sensors exhibit slower response time, typically in milliseconds, compared to
that of the hall-effect sensors and accelerometers. Also, the lifetime of these
sensors is less compared to the hall-effect sensor based technology.
7. Magnetic Sensor Based Technologies
Magnetic sensors [89, 90, 91] are primarily based on the phenomena of
magnetic fields to detect the position of an object and have been widely
used in aerospace, geology and medical sciences [89]. Based on the technolo-
gies used for the magnetic field sensing, the magnetic sensors are of vari-
ous types, such as, search-coil magnetometer [89], optically pumped [92, 93],
fluxgate [94], nuclear precession [90], Superconducting Quantum Interference
Device (SQUlD) [95, 96], magnetodiode [97], magnetoresistive [98], magneto-
transistor [99], fiber-optic magnetometer [89] and magneto-optical [100, 101].
For example, the search-coil magnetometer works on Faradays law of induc-
tion. It consists of a coiled conductor and moving/rotating the coil in the
magnetic field changes its corresponding flux, which generates a voltage be-
tween its leads. The signal detected by a search-coil magnetometer depends
on various factors, such as, number of turns and area of coil, the strength of
magnetic field and the permeability of the material of coil. Table 6 presents
the sensitivity ranges for various magnetic sensors [90].
7.1. Relevant Work
Magnetic sensors have been used for detecting the position of the object
in many applications, such as, aerospace and healthcare. Pabon et al. [102]
presented a data-glove having goniometric sensors. Due to their insensitiv-
ity towards the user’s hand, these sensors do not require any calibration.
Moreover, this data-glove uses the magnetic sensor to track the position and
orientation of the user’s hand. Kortier et al. [103] presented a data glove,
which is based on Inertial and Magnetic Measurement Systems (IMMS) and
provides estimates of the 3D kinematics of fingers and hands. It uses the
magnetic and inertial sensors that are placed on various segments of the
19
Table 6: Sensitivity Ranges for Various Magnetic Sensors
Magnetic Sensors Sensitivity Range
Search-coil magnetometer 2 ×10−5nT, No upper limit
Optically pumped 10−3nT to 105nT
Fluxgate 10−2nT to 107nT
Nuclear precession 10−1nT to 105nT
SQUlD 10 fT or 10−5nT
Magnetoresistive 10−2nT to 103nT
Magnetotransistor 109nT to 1011 nT
Fiber-optic magnetometer 10−2nT to 106nT
hands and fingers and thus enables the accurate measurement of the cor-
responding kinetics. A bidirectional force feedback dataglove is proposed
in [104], which is based on pneumatic artificial muscles. This dataglove uses
the anisotropic magnetoresistive sensors, which provide the measurement of
all the four degree of freedom for each of the fingers such as abduction-
adduction in Metacarpophalangeal (MCP) joint, and flexion and extension
in Distal Interphalangeal (DIP), MCP and Proximal Interphalangeal (PIP)
joints. Due to its light weight and portability, it is widely used in the rehabil-
itation of the hands. A detailed account of the tracking of human motion for
the rehabilitation purposes based on magnetic sensors can be found in [105].
A Pneumatic Glove, namely PneuGlove, and an immersive virtual reality
environment [22] is used for the hand rehabilitation training after stroke.
It uses the magnetic trackers in order to track the position of the head.
Moreover, a flexion sensor is used for the measurement of the joint kinematics
in this environment. Ma et al. [106] proposed a VR based training system,
which is used for the therapy of the stroke patients. It mainly consists of
VR games, which enable a patient with upper limb motor disorders to do
physical exercises and thus serves as a rehabilitation system. In order to
track the movement of the patient’s hand, arms and upper body, it uses four
Ascension MotionStar wireless magnetic sensors [107]. Altun et al. [108] used
the magnetic sensors for the recognition of daily human activities. A detailed
account of the usage of magnetic sensors for human activities recognition and
monitoring, and behavior classification can be found in [109].
20
7.2. Discussion and Analysis
There are many advantages and drawbacks of the magnetic sensors. One
of the main advantage is the variability in their sizes, which make them a
wider utility in many applications. One of the drawbacks of the magnetic
sensors is that their sensitivity changes with their size, which results into
change in power and cost, i.e., it is directly proportional to all these param-
eters. In order to use them for the hand joint monitoring and rehabilitation,
the smaller size, low cost and less power consuming magnetic sensors are pre-
ferred, which lack in their sensing ability as compared to the ones with the
larger size. Moreover, the magnetic sensors exhibit a slower response time
compared to the accelerometers and the flex sensors.
8. An Optimal Rehabilitation Glove
We believe that all the above-mentioned technologies need to play to-
gether in order to develop an optimal rehabilitation glove in terms of accu-
racy, cost and reliability. Table 7 presents a comparison of various wearable
technologies presented in Sections 2 to 7 of this paper based on their accuracy,
performance, cost and the lifetime, which are the most important parameters
while designing a rehabilitation glove. It is important to note that the accu-
racy parameter considers for both the sensing ability and the response time
of a wearable technology, i.e., the desirable accuracy would mean to precisely
and efficiently detect the movements. The green, yellow and red circles rep-
resent the level of behavior of the technologies for each of the parameters.
It can be clearly seen that the flex sensor and accelerometer are the most
optimal technologies based on these parameters since they do not exhibit
any worst ( ) behavior. The flex sensor based technology provides the best
accuracy and lifetime, while the accelerometer based technology provides the
best performance and cost.
21
Table 7: Comparison of Various Wearable Technologies for Rehabilitation Glove
( : Desirable, : Nominal, : Worst)
Technology Accuracy Performance Cost Lifetime
Flex sensor based
Accelerometer based
Vision based
Hall-effect based
Stretch sensor based
Magnetic sensor based
Based on the above-mentioned observations, we propose a wearable re-
habilitation glove, depicted in Figure 7, which utilizes Flexpoint bend sen-
sors [110] to measure the joint angles, the ADXL345 accelerometer [111, 112]
at the wrist and pressure sensors at the fingertips to measure the grip strength
of the hand, since grip is also a major measure of dexterity. We chose the
Flexpoint bend sensors at hand joints due to their decent accuracy, easy cal-
Figure 7: Proposed Glove
ibration and easy placement on the glove. Moreover, these sensors are avail-
able in multiple lengths, which allows us to use them according to the joint
length requirement. An accelerometer is proposed to be used for the wrist
because the wrist movement has two degrees of freedom of movement, which
22
cannot be measured accurately using bend sensors. We also propose to in-
clude a smartphone application, interfaced using Bluetooth, with the glove
for easy user accessibility. The app mainly allows data management of disease
progression and sharing of data with therapists and other patients. More-
over, the app also facilitates the therapists in keeping record of all patients.
This setup is expected to facilitate a comprehensive diagnosis process and
thus more effective treatments.
The major hardware required to implement the proposed wearable in-
cludes Flexpoint bend sensors of lengths 100 and 200 , for DIP joint and for PIP
and MCP joints, respectively. These flex sensors along with the Flexiforce
pressure sensors are placed on a lycra hand glove at the joints and fingertips,
respectively. These flex sensors are calibrated using a calibration setup [113],
which consists of a wooden board, rotating platform and a servo motor. The
wooden board is fixed horizontally on this rotating platform, which is oper-
ated through the servo motor. A metal hinge is placed on the upper end of
the motor’s shaft and the sensor is placed as a cantilever beam on this metal
hinge, as shown in Figure 8. This setup thus allows the sensor to bend at di-
Figure 8: Setup for the Calibration of Flexpoint Bend Sensor
fferent bending rates and angles. For a single degree change in the angle,
the corresponding resistance is measured and the polynomial relationship
acquired from the calibration is then implemented in the microcontroller,
such as ATega32 [114]. The relationship between the angle and resistance
measurements for a 200 unidirectional and a bidirectional flex sensors is de-
picted in Figure 9. Similarly, the Flexiforce A101 [115] pressure sensors are
also used at the joints and fingertips for measuring the values of the pinch
23
pressure of the hand and fingers, respectively. They are made up of polyester
with a linearity error of ±<3%, repeatability of ±<2.5% and a response
time of <5µsec.
0 30 60 90 120 150
Angle (Degree)
0
50
100
150
200
250
300
350
400
450
500
Resistance (Kilo-ohm)
y1 = 0.0336x2 - 1.9309x + 39.093
y2 = 0.0019x3 - 0.2005x2 + 6.134x - 9.8277
Figure 9: Resistance-angle Relationship for a 200 Unidirectional and a Bidirectional Flex
Sensors
During the interfacing of the microcontroller with the sensors, the input
impedance of the microcontroller changes the resistance of the sensors, which
effects the accuracy of the sensors. Thus, to cater far this issue, the raw data
from these sensors is conditioned using a voltage buffer, which is implemented
using the LM324 Operational Amplifiers (Op-Amps) [116]. The flex sensors
are connected to the input of the Op-Amp, whereas, the microcontroller
is connected to the output port of Op-Amp. Thus, this setup caters for
the inaccuracy issue of the sensors caused by the input impedance of the
microcontroller.
Finally, a Bluetooth low Energy (BLE 4.0) [117] module is used to in-
terface the glove with a smartphone application for user accessibility. The
app is used for the record keeping of disease progression and sharing of data
with therapists and other patients. A board with ATMega32 and BLE chip
embedded, like the Blend Micro [118] or RFduino [119], can also be used
for the development of the proposed glove rather than using individual mod-
els for the microcontroller and the Bluetooth. Figure 10 depicts the design
and layout of the android app, whereas Figure 11 shows a prototype of the
developed system.
24
Figure 10: Design and Layout of the Android App
Figure 11: Design of the Proposed Glove
The proposed system exhibited quite promising results for overcoming the
major shortcomings of conventional methods of hand joints monitoring and
thus has the potential to revolutionize the field of hand therapy. The high
accuracy allows us to keep track of even the slightest increase or decrease
in the range of motion of hand joints. The smartphone connectivity allows
sharing the live progress with therapists and other patients, which greatly
25
facilitates the rehabilitation process.
9. Conclusions
In this paper, we presented a survey of all the major wearable technologies
available to measure hand joint angles for rehabilitation process of various
hand deformities. We broadly categorized them in six categories i.e. Flex
sensor based, Accelerometer based, Vision based, Hall-effect based, Stretch
sensor based and Magnetic sensor based. We critically analyzed all these
categories and mentioned their advantages and drawbacks. Finally, based
on various vital parameters, such as, accuracy of sensor’s sensitivity, size,
cost and implementation, we proposed an optimal solution, which provides
a cost-effective, easy and innovative alternative to the current methods of
measuring hand joint angles for the rehabilitation. Our proposed device uses
conductive ink based sensors for finger joints measurement and an accelerom-
eter for the measurement of wrist joint angles. It also uses a smartphone app,
which provides an easy user interfacing and automatic data entry. We also
recommended to use this wearable device to measure the level of dexterity
of a patients hand, which can be done by performing dexterity tests while
donning the wearable and getting necessary readings.
Acknowledgements
We would like to acknowledge the help of our recent Electrical Engineering
graduates, Asad Tariq, Mahnoor Ali and Saad Mahmood, in collecting the
information for this survey and the prototype development of the proposed
optimal rehabilitation glove.
References
[1] U.S National Library of Medicine, https://www.nlm.nih.gov/
medlineplus/handinjuriesanddisorders.html (2017).
[2] J. Condell, K. Curran, T. Quigley, P. Gardiner, M. McNeill, J. Winder,
E. Xie, Z. Qi, J. Connolly, Finger Movement Measurements in Arthritic
Patients using Wearable Sensor Enabled Gloves, International Journal
of Human Factors Modelling and Simulation 2 (4) (2011) 276–292.
26
[3] I. Ibrahim, W. Khan, N. Goddard, P. Smitham, Carpal Tunnel Syn-
drome: A Review of the Recent Literature, The Open Orthopaedics
Journal 6 (2012) 69–76.
[4] OT-Hand Therapy, https://www.pinterest.com/corinnevisco/
ot-hand-therapy/ (2017).
[5] R. M. d. Carvalho, N. Mazzer, C. H. Barbieri, Analysis of the Reliabil-
ity and Reproducibility of Goniometry Compared to Hand Photogram-
metry, Acta Ortop Brasileira 20 (2012) 139 – 149.
[6] N. M. Massy-Westropp, T. K. Gill, A. W. Taylor, R. W. Bohannon,
C. L. Hill, Hand Grip Strength: Age and Gender Stratified Normative
Data in a Population-based Study, BMC Research Notes 4 (1) (2011)
127.
[7] S. V. Duff, D. H. Aaron, G. R. Gogola, F. J. Valero-Cuevas, Innovative
Evaluation of Dexterity in Pediatrics, Journal of Hand Therapy 28 (2)
(2015) 144–150.
[8] G. Saggio, F. Riillo, L. Sbernini, L. R. Quitadamo, Resistive Flex Sen-
sors: A Survey, Smart Materials and Structures 25 (1) (2015) 013001.
[9] Buckling of Bend Sensors in Fabric Tubes, http://dev-blog.
mimugloves.com/buckling-of-bend-sensors-in-fabric-tubes/
(2017).
[10] Bend Sensor Technology Electronic Interface Guide, http://www.
flexpoint.com/media-resources/electrical-data-sheets/
(2017).
[11] Spectra Symbol Flex Sensor, https://www.sparkfun.com/
datasheets/Sensors/Flex/flex22.pdf (2017).
[12] CyberGlove Systems, http://www.cyberglovesystems.com/
cyberglove-iii/ (2017).
[13] P. Kumar, J. Verma, S. Prasad, Hand Data Glove: A Wearable Real-
time Device for Human-computer Interaction, International Journal of
Advanced Science and Technology 43 (2012) 15–25.
27
[14] J. Connolly, K. Curran, J. Condell, P. Gardiner, Wearable Rehab Tech-
nology for Automatic Measurement of Patients with Arthritis, in: Per-
vasive Computing Technologies for Healthcare, 2011, pp. 508–509.
[15] B. O’Flynn, J. T. Sanchez, P. Angove, J. Connolly, J. Condell, K. Cur-
ran, P. Gardiner, Novel Smart Sensor Glove for Arthritis Rehabiliation,
in: Body Sensor Networks, 2013, pp. 1–6.
[16] L. Gallo, A Glove-based Interface for 3D Medical Image Visualization,
in: Intelligent Interactive Multimedia Systems and Services, Springer,
2010, pp. 221–230.
[17] D. VHand, 2.0 OEM Technical Datasheet, Tech. rep., DGTech Engi-
neering Solutions (2007).
[18] T. G. Zimmerman, J. Lanier, C. Blanchard, S. Bryson, Y. Harvill, A
Hand Gesture Interface Device, in: ACM SIGCHI Bulletin, Vol. 18,
ACM, 1987, pp. 189–192.
[19] L. Simone, E. Elovic, U. Kalambur, D. Kamper, A Low Cost Method to
Measure Finger Flexion in Individuals with Reduced Hand and Finger
Range of Motion, in: Engineering in Medicine and Biology Society,
Vol. 2, IEEE, 2004, pp. 4791–4794.
[20] G. Saggio, A Novel Array of Flex Sensors for a Goniometric Glove,
Sensors and Actuators A: Physical 205 (2014) 119–125.
[21] D. Zhang, Y. Shen, S.-K. Ong, A. Y. Nee, An Affordable Augmented
Reality based Rehabilitation System for Hand Motions, in: Cyber-
worlds, IEEE, 2010, pp. 346–353.
[22] L. Connelly, Y. Jia, M. L. Toro, M. E. Stoykov, R. V. Kenyon, D. G.
Kamper, A Pneumatic Glove and Immersive Virtual Reality Environ-
ment for Hand Rehabilitative Training after Stroke, IEEE Transactions
on Neural Systems and Rehabilitation Engineering 18 (5) (2010) 551–
559.
[23] L. Dipietro, A. M. Sabatini, P. Dario, A Survey of Glove-based Sys-
tems and their Applications, IEEE Transactions on Systems, Man, and
Cybernetics, Part C (Applications and Reviews) 38 (4) (2008) 461–482.
28
[24] Flex Sensor, http://www.sensorwiki.org/doku.php/sensors/
flexion (2017).
[25] B. B. Graham, Using an Accelerometer Sensor to Measure Human
Hand Motion, Ph.D. thesis, Massachusetts Institute of Technology
(2000).
[26] J. S. Wilson, Sensor Technology Handbook, Elsevier, 2004.
[27] C. P. B. Gole, A. Banmare, Wearable Computing Using Key glove,
in: Recent Advances in Technology and Management for Integrated
Growth, 2013, pp. 329–336.
[28] P. C. Hsiao, S. Y. Yang, B. S. Lin, I. J. Lee, W. Chou, Data Glove
Embedded with 9-axis IMU and Force Sensing Sensors for Evaluation
of Hand Function, in: Engineering in Medicine and Biology Society,
2015, pp. 4631–4634.
[29] A. Heinrich, Gest Glove has Gesture Control on Hand, http://www.
gizmag.com/gest-gesture-controller-glove/40174/ (2017).
[30] K. Ellis, J. C. Barca, Exploring Sensor Gloves for Teaching Children
Sign Language, Advances in Human-Computer Interaction (2012) 12.
[31] KeyGlove, Freedom in the Palm of Your Hand, http://www.keyglove.
net/ (2017).
[32] T. D. Bui, L. T. Nguyen, Recognizing Postures in Vietnamese Sign
Language with MEMS Accelerometers, IEEE Sensors Journal 7 (5)
(2007) 707–712.
[33] J.-H. Kim, N. D. Thang, T.-S. Kim, 3-D Hand Motion Tracking and
Gesture Recognition Using a Data Glove, in: International Symposium
on Industrial Electronics, IEEE, 2009, pp. 1013–1018.
[34] R. Xu, S. Zhou, W. J. Li, MEMS Accelerometer Based Nonspecific-
user Hand Gesture Recognition, IEEE Sensors Journal 12 (5) (2012)
1166–1173.
[35] J. L. Hernandez-Rebollar, E. I. Elsakay, J. D. Alan´ıs-Urquieta, Ac-
celespell, a Gestural Interactive Game to Learn and Practice Finger
Spelling, in: Multimodal Interfaces, ACM, 2008, pp. 189–190.
29
[36] B. OFlynn, J. T. Sanchez, J. Connolly, J. Condell, K. Curran, P. Gar-
diner, B. Downes, Integrated Smart Glove for Hand Motion Monitor-
ing, in: Sensor Device Technologies and Applications, 2015.
[37] C.-C. Yang, Y.-L. Hsu, A Review of Accelerometry-based Wearable
Motion Detectors for Physical Activity Monitoring, Sensors 10 (8)
(2010) 7772–7788.
[38] S. Patel, H. Park, P. Bonato, L. Chan, M. Rodgers, A Review of Wear-
able Sensors and Systems with Application in Rehabilitation, Journal
of Neuroengineering and Rehabilitation 9 (1) (2012) 21.
[39] KC-2105 Accelerometer, https://www.kineticceramics.com/
accelerometers (2017).
[40] P. Garg, N. Aggarwal, S. Sofat, Vision Based Hand Gesture Recogni-
tion, World Academy of Science, Engineering and Technology 49 (1)
(2009) 972–977.
[41] R. R. Itkarkar, A. K. Nandy, A Study of Vision Based Hand Gesture
Recognition for Human Machine Interaction, Innovative Research in
Advanced Engineering (2014) 2349–2163.
[42] J. Singha, K. Das, Hand Gesture Recognition Based on Karhunen-
Loeve Transform, arXiv preprint arXiv:1306.2599.
[43] C. W. Chang, C. H. Chang, A Two-hand Multi-point Gesture Recog-
nition System Based on Adaptive Skin Color Model, in: Consumer
Electronics, Communications and Networks, 2011, pp. 2901–2904.
[44] T. Kapuscinski, M. Wysocki, Hand Gesture Recognition for Man-
machine Interaction, in: Robot Motion and Control, 2001, pp. 91–96.
[45] C. Yu, X. Wang, H. Huang, J. Shen, K. Wu, Vision-based Hand Gesture
Recognition using Combinational Features, in: Intelligent Information
Hiding and Multimedia Signal Processing, 2010, pp. 543–546.
[46] A. Malima, E. Ozgur, M. C¸ etin, A Fast Algorithm for Vision-based
Hand Gesture Recognition for Robot Control, in: Signal Processing
and Communications Applications, 2006, pp. 1–4.
30
[47] M. Manigandan, I. M. Jackin, Wireless Vision Based Mobile Robot
Control Using Hand Gesture Recognition Through Perceptual Color
Space, in: Advances in Computer Engineering, 2010, pp. 95–99.
[48] E. Koh, J. Won, C. Bae, On-premise Skin Color Modeing Method for
Vision-based Hand Tracking, in: Consumer Electronics, 2009, pp. 908–
909.
[49] Y. Fang, K. Wang, J. Cheng, H. Lu, A Real-time Hand Gesture Recog-
nition Method, in: Multimedia and Expo, 2007, pp. 995–998.
[50] J. Rekha, J. Bhattacharya, S. Majumder, Shape, Texture and Local
Movement Hand Gesture Features for Indian Sign Language Recog-
nition, in: Trendz in Information Sciences & Computing, 2011, pp.
30–35.
[51] X. Zabulis, H. Baltzakis, A. Argyros, Vision-based Hand Gesture
Recognition for Human-computer Interaction, in: The Universal Ac-
cess Handbook, CRC Press, 2009, pp. 1–30.
[52] M. S. Cameir˜ao, S. B. i Badia, L. Zimmerli, E. D. Oller, P. F. Verschure,
The Rehabilitation Gaming System: A Virtual Reality based System
for the Evaluation and Rehabilitation of Motor Deficits, in: Virtual
Rehabilitation, IEEE, 2007, pp. 29–33.
[53] G. Placidi, D. Avola, D. Iacoviello, L. Cinque, Overall Design and Im-
plementation of the Virtual Glove, Computers in Biology and Medicine
43 (11) (2013) 1927–1940.
[54] Z. Ma, P. Ben-Tzvi, J. Danoff, Modeling Human Hand and Sens-
ing Hand Motions with the Five-fingered Haptic Glove Mechanism,
in: ASME 2014 International Design Engineering Technical Confer-
ences and Computers and Information in Engineering Conference,
American Society of Mechanical Engineers, 2014, pp. V05AT08A008–
V05AT08A008.
[55] J. P. Wachs, M. K¨olsch, H. Stern, Y. Edan, Vision-based Hand-gesture
Applications, Communications of the ACM 54 (2) (2011) 60–71.
31
[56] B. Baatar, J. Tanaka, Comparing Sensor Based and Vision Based Tech-
niques for Dynamic Gesture Recognition, in: Computer Human Inter-
action, 2012.
[57] R. Y. Wang, J. Popovi´c, Real-time Hand-tracking with a Color Glove,
in: ACM Transactions on Graphics, Vol. 28, 2009, pp. 63–69.
[58] E. Ramsden, Hall-effect Sensors: Theory and Application, Newnes,
2011.
[59] R. S. Popovic, Hall Effect Devices, CRC Press, 2003.
[60] L. Dipietro, A. M. Sabatini, P. Dario, Evaluation of an Instrumented
Glove for Hand-movement Acquisition, Rehabilitation Research and
Development 40 (2) (2003) 179–190.
[61] Humanglove by Humanware SRL, http://www.hmw.it (1997).
[62] Humanglove Developers Manual, Humanware, Pisa, Italy (1998).
[63] I. Sarakoglou, N. G. Tsagarakis, D. G. Caldwell, Occupational and
Physical Therapy using a Hand Exoskeleton Based Exerciser, in: In-
telligent Robots and Systems, Vol. 3, IEEE, 2004, pp. 2973–2978.
[64] J. Wu, J. Huang, Y. Wang, K. Xing, RLSESN-based PID Adaptive
Control for a Novel Wearable Rehabilitation Robotic Hand Driven by
PM-TS Actuators, International Journal of Intelligent Computing and
Cybernetics 5 (1) (2012) 91–110.
[65] M. Carrozza, B. Massa, S. Micera, M. Zecca, P. Dario, A Wearable
Artificial Hand for Prosthetics and Humanoid Robotics Applications,
in: IEEE-RAS International Conference on Humanoid Robots, 2001.
[66] Honeywell Sensors, https://sensing.honeywell.com/
SS495A-S-linear-and-angle-sensor-ics (2017).
[67] L. Chan, R.-H. Liang, M.-C. Tsai, K.-Y. Cheng, C.-H. Su, M. Y. Chen,
W.-H. Cheng, B.-Y. Chen, FingerPad: Private and Subtle Interaction
using Fingertips, in: Symposium on User interface Software and Tech-
nology, ACM, 2013, pp. 255–260.
32
[68] G. Phillips, D. McGrouther, B. Andrews, Finger Mobility Following
Flexor Tendon Repair, Hand Surgery (British and European Volume)
10 (3) (1985) 337–339.
[69] T. Chouhan, A. Panse, A. K. Voona, S. Sameer, Smart Glove with
Gesture Recognition Ability for the Hearing and Speech Impaired,
in: Global Humanitarian Technology Conference-South Asia Satellite,
IEEE, 2014, pp. 105–110.
[70] M. F. Eilenberg, H. Geyer, H. Herr, Control of a Powered Ankle-foot
Prosthesis Based on a Neuromuscular Model, IEEE Transactions on
Neural Systems and Rehabilitation Engineering 18 (2) (2010) 164–173.
[71] Allegro MicroSystems, http://www.allegromicro.com/en/
Products/Magnetic-Linear-And-Angular-Position-Sensor-ICs/
Linear-Position-Sensor-ICs/A1391-2-3-5.aspx (2017).
[72] A. Arami, N. V. Martins, K. Aminian, Locally Linear Neuro-Fuzzy
Estimate of the Prosthetic Knee Angle and Its Validation in a Robotic
Simulator, IEEE Sensors Journal 15 (11) (2015) 6271–6278.
[73] VF360NT Hall-effect Sensor, http://www.mouser.com/ds/2/187/
honeywell-sensing-hall-effect-sensor-ics-vf360nt-v-775784.
pdf (2017).
[74] Z. Shen, J. Yi, X. Li, M. H. P. Lo, M. Z. Chen, Y. Hu, Z. Wang, A Soft
Stretchable Bending Sensor and Data Glove Applications, Robotics
and Biomimetics 3 (1) (2016) 22.
[75] Stretch Sensors, http://www.imagesco.com/sensors/stretch.pdf
(2017).
[76] H. Kim, S. Park, N. Na, J. Kim, Y. Moon, J. Kim, The Smart Arm-
band: Expanding Wearable Interface Area and Suggesting Interaction
Scenarios, in: Information Science and Applications, Springer, 2016,
pp. 1361–1365.
[77] Fabric Stretch Sensors, http://www.kobakant.at/DIY/?p=210
(2017).
33
[78] Constructed Stretch Sensors, http://www.kobakant.at/DIY/?p=1781
(2017).
[79] Knit Stretch Sensors, http://www.kobakant.at/DIY/?p=1762 (2017).
[80] Circular Knit Stretch Sensors, http://www.kobakant.at/DIY/?p=
2108 (2017).
[81] A. Toomey, R. Oliver, N. OConnor, P. Stevenson-Keating, A Design-
led, Materials Based Approach to Human Centered Applications using
Modified Dielectric Electroactive Polymer Sensors, in: Sensor Systems
and Software, Springer, 2014, pp. 11–19.
[82] G. B¨uscher, R. K˜oiva, C. Sch¨urmann, R. Haschke, H. J. Ritter, Tac-
tile Dataglove with Fabric-based Sensors, in: Humanoid Robots (Hu-
manoids), IEEE, 2012, pp. 204–209.
[83] M. Bianchi, R. Haschke, G. B¨uscher, S. Ciotti, N. Carbonaro,
A. Tognetti, A Multi-Modal Sensing Glove for Human Manual-
Interaction Studies, Electronics 5 (3) (2016) 42–56.
[84] T.-Y. Chuang, W.-S. Huang, S.-C. Chiang, Y.-A. Tsai, J.-L. Doong,
H. Cheng, A Virtual Reality-based System for Hand Function Analysis,
Computer Methods and Programs in Biomedicine 69 (3) (2002) 189–
196.
[85] J. Lee, S. Kim, J. Lee, D. Yang, B. C. Park, S. Ryu, I. Park, A Stretch-
able Strain Sensor Based on a Metal Nanoparticle Thin Film for Human
Motion Detection, Nanoscale 6 (20) (2014) 11932–11939.
[86] A. Tognetti, F. Lorussi, G. Dalle Mura, N. Carbonaro, M. Pacelli,
R. Paradiso, D. De Rossi, New Generation of Wearable Goniometers
for Motion Capture Systems, Journal of Neuroengineering and Reha-
bilitation 11 (1) (2014) 56–72.
[87] Y. Shimada, K. Sato, T. Matsunaga, Y. Tsutsumi, A. Misawa, S. Ando,
T. Minato, M. Sato, S. Chida, K. Hatakeyama, Closed-loop Control
using a Stretch Sensor for Restoration of Standing with Functional
Electrical Stimulation in Complete Paraplegia, The Tohoku Journal of
Experimental Medicine 193 (3) (2001) 221–227.
34
[88] R. McLaren, F. Joseph, C. Baguley, D. Taylor, A Review of E-textiles
in Neurological Rehabilitation: How Close are We?, Journal of Neuro-
Engineering and Rehabilitation 13 (1) (2016) 59.
[89] J. Lenz, S. Edelstein, Magnetic Sensors and their Applications, IEEE
Sensors Journal 6 (3) (2006) 631–649.
[90] J. E. Lenz, A Review of Magnetic Sensors, Proceedings of the IEEE
78 (6) (1990) 973–989.
[91] P. Ripka, Magnetic Sensors and Magnetometers, Artech House, 2001.
[92] W. Happer, Optical Pumping, Reviews of Modern Physics 44 (2) (1972)
169.
[93] D. Budker, W. Gawlik, D. Kimball, S. Rochester, V. Yashchuk,
A. Weis, Resonant Nonlinear Magneto-optical Effects in Atoms, Re-
views of Modern Physics 74 (4) (2002) 1153.
[94] P. Ripka, Advances in Fluxgate Sensors, Sensors and Actuators A:
Physical 106 (1) (2003) 8–14.
[95] V. Pizzella, S. Della Penna, C. Del Gratta, G. L. Romani, SQUID
Systems for Biomagnetic Imaging, Superconductor Science and Tech-
nology 14 (7) (2001) R79.
[96] R. Cantor, SQUIDS And Emerging Applications, Superconductor and
Cryoelectronics 13 (4) (2000) 16–22.
[97] R. Popovic, H. P. Baltes, F. Rudolf, An Integrated Silicon Magnetic
Field Sensor Using the Magnetodiode Principle, IEEE Transactions on
Electron Devices 31 (3) (1984) 286–291.
[98] S. Tumanski, Thin Film Magnetoresistive Sensors, CRC Press, 2001.
[99] C. S. Roumenin, Bipolar Magnetotransistor Sensors. An Invited Re-
view, Sensors and Actuators A: Physical 24 (2) (1990) 83–105.
[100] P. Zu, C. C. Chan, W. S. Lew, Y. Jin, Y. Zhang, H. F. Liew, L. H.
Chen, W. C. Wong, X. Dong, Magneto-optical Fiber Sensor Based on
Magnetic Fluid, Optics letters 37 (3) (2012) 398–400.
35
[101] P. Zu, C. Chiu Chan, T. Gong, Y. Jin, W. Chang Wong, X. Dong,
Magneto-optical Fiber Sensor Based on Bandgap Effect of Photonic
Crystal Fiber Infiltrated with Magnetic Fluid, Applied Physics Letters
101 (24) (2012) 241118.
[102] S. Pabon, E. Sotgiu, R. Leonardi, C. Brancolini, O. Portillo-Rodriguez,
A. Frisoli, M. Bergamasco, A Data-glove with Vibro-tactile Stimula-
tors for Virtual Social Interaction and Rehabilitation, in: International
Workshop on Presence, 2007, pp. 25–27.
[103] H. G. Kortier, V. I. Sluiter, D. Roetenberg, P. H. Veltink, Assessment
of Hand Kinematics using Inertial and Magnetic Sensors, Neuroengi-
neering and Rehabilitation 11 (1) (2014) 1–14.
[104] Z. Sun, X. Miao, X. Li, Design of a Bidirectional Force Feedback Data-
glove based on Pneumatic Artificial Muscles, in: Mechatronics and
Automation, IEEE, 2009, pp. 1767–1771.
[105] H. Zhou, H. Hu, Human Motion Tracking for Rehabilitation–A Survey,
Biomedical Signal Processing and Control 3 (1) (2008) 1–18.
[106] M. Ma, M. McNeill, D. Charles, S. McDonough, J. Crosbie, L. Oliver,
C. McGoldrick, Adaptive Virtual Reality Games for Rehabilitation of
Motor Disorders, Universal Access in Human-Computer Interaction.
Ambient Interaction (2007) 681–690.
[107] MotionStar wireless, http://www.mindflux.com.au/products/
ascension/motionstar-wireless.html (2017).
[108] K. Altun, B. Barshan, Human Activity Recognition using Iner-
tial/Magnetic Sensor Units, in: International Workshop on Human
Behavior Understanding, Springer, 2010, pp. 38–51.
[109] Z. Wang, Z. Yang, T. Dong, A Review of Wearable Technologies for El-
derly Care that Can Accurately Track Indoor Position, Recognize Phys-
ical Activities and Monitor Vital Signs in Real Time, Sensors 17 (2)
(2017) 341.
[110] Flexpoint Bend Sensor, http://www.sensorwiki.org/doku.php/
sensors/flexion (2017).
36
[111] The ADXL345 Accelerometer, http://www.analog.com/en/
products/mems/accelerometers/adxl345.html#product-overview
(2017).
[112] ADXL345 Digital Accelerometer, https://cdn-learn.adafruit.
com/downloads/pdf/adxl345-digital-accelerometer.pdf (2017).
[113] G. Orengo, G. Saggio, S. Bocchetti, F. Giannini, Advanced Character-
ization of Piezoresistive Sensors for Human Body Movement Tracking,
in: International Symposium on Circuits and Systems, IEEE, 2010, pp.
1181–1184.
[114] ATmega32 Microcontroller, http://www.atmel.com/devices/
atmega32.aspx (2017).
[115] FlexiForce A101 Sensor, https://www.tekscan.com/
products-solutions/force-sensors/a101 (2017).
[116] LM324, http://www.onsemi.com/pub_link/Collateral/LM324-D.
PDF (2017).
[117] J. Decuir, et al., Bluetooth 4.0: Low Energy, Cambridge, UK: Cam-
bridge Silicon Radio SR PLC 16, 2010.
[118] Blend Micro, http://redbearlab.com/blendmicro/ (2017).
[119] RFduino, http://www.rfduino.com/ (2017).
37