A soft stretchable bending sensor and data glove applications
Shen et al. Robot. Biomim.
A soft stretchable bending sensor and data glove applications
Zhong Shen 0
Juan Yi 0 2
Xiaodong Li 1
Mark Hin Pei Lo 1
Michael Z. Q. Chen 0 2
Yong Hu 1
Zheng Wang 0 2
0 Department of Mechanical Engineering, The University of Hong Kong , Pok Fu Lam, Hong Kong SAR , China
1 Department of Orthopaedics and Traumatology, The University of Hong Kong , Pok Fu Lam, Hong Kong SAR , China
2 HKU Shenzhen Institute of Research and Inno- vation , Shenzhen , China
Soft sensors are required to accommodate the flexible and deformable natures of the human body in wearable device applications. They are also suitable for integration with soft robotic devices to monitor the performance status and provide references for feedback control. However, the choices for bending sensors are still highly limited. In this paper, a soft bending sensor is presented. By careful design with a blend of sensitive and insensitive regions, the sensor could be stretchable while being insensitive to stretching. An analytical study was presented on how to design the sensor with the named bending/stretching feature. This feature enables the sensor to be implemented in measuring human motions where a large amount of skin stretch is involved. Two sensor gloves were designed and fabricated based on the proposed soft bending sensor, aiming for different application scenarios. Both the sensor and the gloves were evaluated using a dedicated evaluation platform with experimental results compared against each other.
Soft sensors are receiving growing attention, due to both
the global wave of developments in wearable
human-centered devices and the recent focus on soft robots [1–4].
For human motions, soft sensors could provide direct
joint-level angle measurements, hence lead to a
reconstruction of human body trajectories. For soft robots,
soft sensors are required to provide sensory and
control information while not interfering with the primary
compliant and adaptive features of the robotic devices.
However, for both of the above application scenarios,
bending and stretch are two closely coupled factors. It is
technically very challenging to make stretchable bending
sensors insensitive to stretching. The available flexible
bending sensors are very limited, with few allowing for
In the last two decades, many researchers have engaged
in developing a new kind of sensing technology, whether
in hardware or software, to meet the need of soft
wearable devices. Optical fiber sensors, which have been
recognized as a standard for motion capture, have drawn some
attention [5–7]. Apart from optical fiber sensors, there
were other groups making similar wearable motion
capture devices using resistive bending sensors [8–10] and
using a filter and multiple sensors for one joint to reduce
the error . Except resistive sensors, printed sensors or
MEMS sensors can also provide small size, low cost and
flexibility [12, 13]. Recently, a new kind of sensors that
uses a conductive liquid metal injected in a soft chamber
has drawn much attention. The resistance of the sensor
will change as the shape of the chamber deforms due to
external force [14–18]. This soft stretchable bending
sensor suits well for wearable devices, but the fluidity of
liquid metal still brings technical constraints and limits its
The techniques mentioned above have contributed to
the wearable sensing technologies. For wider and further
applications, bending or strain sensors are still in need of
flexibility, comfort and accuracy as well as low cost and
non-toxicity. So, here we present a novel low-cost soft
stretchable bending sensor (Fig. 1) along with two
different sensor gloves. Both the sensor and the gloves have
excellent flexibility as well as precision. The sizes of the
sensor and the gloves are also customizable, which is a
key issue for wearable devices.
In “Sensor and sensor glove” section, the design details
of the bending sensor and two kinds of sensor gloves
© The Author(s) 2016. This article is distributed under the terms of the Creative Commons Attribution 4.0 International License
(http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium,
provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license,
and indicate if changes were made.
three parts: two belts and the middle part (Fig. 2).
Considering the length and elongation of the sensor:
Fig. 1 Illustration of a soft stretchable bending sensor mounted on
an adult’s hand. The sensor uses EPR embedded in a soft stretchable
base. The soft and stretch abilities of the sensor allow it to attach
firmly on the finger joint
are described. “Sensor evaluation” section illustrates the
experimental methods and results for the bending
sensor. “Glove evaluation” section shows the testing results
for two kinds of sensor gloves. “Conclusions” section
presents the conclusion and future work.
Sensor and sensor glove
Bending fingers means stretched surface, but most
existing sensors are non-stretchable. The sensor’s slipping on
skin may cause serious problems and increase error. In
this case, we design a stretchable bending sensor that is
insensitive to stretching.
The sensors are based on ethylene propylene rubber
(EPR). We used the Scotch Electrical Semi-Conducting
Tape 13 (3M Company, Maplewood, Minnesota, USA)
as the sensing material. When the material is bent or
stretched, the resistance will change. Typical physical and
electrical properties of the material are shown in Table 1.
EPR will have elastic deformation when elongation is less
than 2% , but the elongation of stretched skin may
reach 40%, which may cause serious plastic deformation
and change the resistance permanently. In order to make
the sensor stretchable, a structure is needed to
separate stretch from bending. We settled on a soft housing
for the EPR portion. The whole structure is consisted of
Table 1 Typical properties of the sensing material
l = l1 + l2 + l3
Ehl = E1l1 + E2l2 + E3l3
f = aiEiliEi (i = 1, 2, 3)
l is the length of the sensor in (1) and (3), l is the
elongation and Eh is the maximum strain caused by finger
bend. Using Hooke’s law, we have:
f is the force applied on the sensor, a is the
cross-sectional area and E is the Young’s modulus. Thus,
Substituting (5) into (3), we have:
Eq. (7) provides a method of determining the geometric
design of the sensor to ensure the elongation of the
sensor section is always within the repeatable region.
A soft rubber called Ecoflex 00-30(Smooth-On, East
Texas, PA 18046) was used to mold the sensor, and three
steps are needed to fabricate a sensor. First the main
structure at the bottom is molded. Then, the sensing
material is put in the middle of the structure and
connected to the circuit. In order to maintain the soft
ability of the sensor, a sliver-plated nylon thread(Less EMF
Inc, Latham, NY, USA, Cat.#A1226) is used to connect
the sensor to the circuit. As shown in Fig. 3, the thread is
first sewed into the two distal end of the sensor and then
sewed into the soft belts. Because the nylon thread is not
stretchable, we reserved some length in the soft belts to
prevent conductive thread from breaking when the belts
were stretched. After the connection process, a coating
layer is added on the sensing material to seal up the
0.16 g cm−3
Fig. 2 a Sensor structure. The middle part is designed to be thicker
and wider than the two belts, offering better stiffness to avoid
stretching. b Profile chart of the sensor. The middle part, which
contains the sensing material, will bend with permitted stretch.
Elongation caused by finger bend will be offset by the stretch of the
Fabric data‑collecting glove
Hand posture is one of the most representative and
complicated motion for human body motion capture. Here,
we present a sensor glove for monitoring hand posture as
well as testing the sensor (Fig. 4).
Most of the existing bending sensors are
non-stretchable, and they will slip when fingers bend. So traditional
data gloves have to be made larger than the hand size
to reserve some space for the sensors to slip. The gap
between the glove and hand may cause some significant
error. For the fabric data-collecting glove, the glove size
can be fully customized due to the stretch ability of the
EPR-based bending sensor. Each sensor can be fixed on
the fabric glove without slipping so the error caused by
sensor slipping can be minimized.
In sum, there are 10 bending sensors used in the sensor
glove and all the resistance are measured using Arduino
Mega 2560 in the white box. Two sensors are used on the
thumb to measure metacarpophalangeal joint (MP) and
distal interphalangeal joint (DIP). Except for the thumb,
two sensors are used for each finger to measure the
bending angle of MP joint and proximal interphalangeal (PIP)
joint; DIP joints are not measured. Since there is a
linkage between the proximal and distal interphalangeal
joints, DIP joints cannot bend independently .
Soft rubber data‑collecting glove
Most existing data-collecting gloves are designed and
fabricated based on traditional fabric gloves. The size
of the glove is fixed, and normally the glove does not fit
well for different individuals and the complex structure
of data gloves will affect the precision. In order to avoid
these problems, a new modular data-collecting glove is
developed for hand posture capture (Fig. 5).
Fig. 3 Connection method for EPR sensor. A sliver-plated nylon
thread is used to connect the sensor to the circuit. A certain length
has been reserved in the soft belts to prevent the thread from
breaking when the belts are stretched
Fig. 4 Fabric data-collecting glove. In total 10 soft stretchable
sensors were integrated in the glove. The white box, which contains the
electrical components, is used to connect the glove to the computer
Fig. 5 Soft rubber data-collecting glove. The top of each part of the
glove is a ring-shaped structure, and it is used to locate one end of
the glove on the fingertips. The bottom of the five parts fixed inside
the hook&loop to locate the other end of the glove on the wrist.
There is a preload in each part of the glove to keep the sensing zones
on the corresponding joints
The novel data-collecting glove uses soft rubber as the
base material instead of traditional fabric gloves. The soft
stretchable structure makes it possible for providing the
same action like the skin. The glove is made of five parts
representing five fingers, and all parts have the same
structure. For each part, there are two sensing zones and
three connecting belts. The belt in the middle connects
the two sensing zones, and the length of the belt is
various to fit in with different finger sizes. The other belts
are used to locate the glove on a hand. Due to the stretch
ability of the soft rubber, the glove can be easily adjusted
to adapt to different hand sizes and each sensor can be
precisely mounted on the corresponding finger joint. In
fact, this method can be used for multiple joints
measurement. Assuming that for each part there are n
sensing zones and n + 1 connecting belts, the total elongation
E L =
(EsiLsi + EciLci) + Ec(n+1)Lc(n+1)
L is the length of each part and E is the strain of each
part. Using Hooke’s law, we have:
(i = 1, 2, 3, . . . , n),
A is the cross-sectional area and E is the Young’s
modulus. Substituting (9) into (8),
E L −
Eq. (10) provides a method of determining the
geometric design of each sensing zone to ensure the
elongation of the sensor section is always within the repeatable
region. Compared with traditional data gloves, this glove
acts just like the skin on hand so it can directly
measure the bending motion without any intermediary. And
the soft modular design of the glove greatly expands the
scope of applications such as nuclear magnetic resonance
imaging(MRI) compatible scenarios.
Bending fingers always comes up with two portions: the
bending motion and the skin stretch. Either portion can
be regarded as a measurement object. Thus, the
sensing material’s sensitivity to bend and stretch is tested
separately (Fig. 6). Three different sizes with respect to
length-to-width ratio are designed to explore the
relationship between sensor size and sensitivity.
The results illustrate that the sensor exhibits
excellent sensitivity to both bend and stretch. For bending
motion, the resistance changing ranges for the size of
5 × 20 mm, 10 × 30 mm and 10 × 20 mm are 30.6, 29.2
and 33.52%, respectively. All the three designs show
similar sensitivity to bend, and the sensitivity to bend is
independent to sensor size. For stretching, the sensor of size
10 × 30 mm and 10 × 20 mm exhibit similar sensitivity
(86.0 and 79.8%) and the sensor of size 5 × 20 mm has a
resistance changing range of 67.1%, illustrating the
sensitivity to stretch is related to the width of the sensor.
The sensing material is sensitive to both bend and stretch.
But stretching will cause unrecoverable plastic
deformation, and the resistance will change permanently as well.
By using the soft structure mentioned in “Sensor and
sensor glove” section, the sensing material’s stretching is
avoided and repeatability of bending is tested. In order to
test the sensor’s repeatability apart from the soft
structure, a dedicated evaluation platform was developed. The
Fig. 6 Testing results of the sensor’s sensitivity to bend and stretch.
a Sensitivity to bend. b Sensitivity to stretch. Unrecoverable plastic
deformation will take place when elongation is up to 2%. Three
different sizes are designed for both tests, and their width-to-length ratio is
all set to 1:2, 1:3 and 1:4 for comparison
platform has two-degrees-of-freedom motion in the
horizontal and vertical movements. To test the repeatability
of the sensor, a highly repeatable and precise testing
procedure is in need. The procedure should also be simple to
minimize the inherent error. In this case, we produce a
specific structure to test the sensor (Fig. 7).
The structure consists of 6 parts. Sensor is the black
part. There are notches on Part 3 and Part 4, which are
used to fix the sensor’s two edges. The rotation of these
two parts can avoid bending on two edges. The sensor is
flat in the beginning. Assume the length of the sensor is
l. When Part 6 moves l, the sensor will be bent in the
middle. Consider the bent sensor as an arc, the radian
should be the bending angle of the sensor. As the length
of the sensor is a constant, we have:
Fig. 7 a Testing platform. Both axes are driven by stepper motor,
and the minimum displacement is 1 mm. b Testing structure used to
transform linear movement into bending motion. The black part is
the sensor separated from the stretchable structure
The bending angle of the sensor can be obtained by
Three different sizes of sensors are designed to access
the influence of sensor size on repeatability, and the sizes
of sensors are 10 × 20 mm, 10 × 30 mm and 5 × 20 mm
respectively. Figure 8 shows the testing results. The
average error between the standard value and the
measured value are 1.87, 10.53 and 1.46% for the size of
10 × 20 mm, 10 × 30 mm and 5 × 20 mm respectively.
The sensors with the size of 10 × 20 mm and 5 × 20 mm
exhibit good repeatability, while the sensor with the size
of 10 × 30 mm has a larger error compared with the other
two sensors, proving that when the length of the sensor
exceeds 20 mm, the repeatability will drop.
Fig. 8 Testing results of repeatability for three different sizes of
sensors, illustrating good repeatability for the size of10 × 20 mm and
5 × 20 mm. The testing results for the size of10 × 30 mm illustrate
that the length of the sensor should not exceed 20 mm
According to the sensitivity test and repeatability test, the
sensing material with the size of 5 mm × 20 mm
exhibited the best performance. In this case, we used the size
5 mm × 20 mm to test the drifting. we used the same
platform in the last subsection to test the drifting. The
sensor was flat at the very beginning. After 30 s, the
sensor was bent to 60◦ and waited for 30 s. Then, the sensor
was bent to 120◦ and waited for 30 s. Then, the sensor
was flat again and the whole process was repeated for 5
times. The testing result is shown in Fig. 9.
The testing results show that the sensing material’s
resistance will drop when the material is kept in a
certain angle. The decrease in resistance is 8.5, 3.1 and 1.5%
when then bending angle is 0◦, 60◦ and 120◦ respectively.
Fabric data‑collecting glove
In order to evaluate the repeatability and precision of
fabric data-collecting glove, a commercially available bending
sensor, the Flex Sensor 2.2 (Spectra Symbol, Salt Lake City,
UT 84119, USA), is used as a contrast to test the fabric
sensor glove. Figure 10 exhibits the evaluation of Flex
Sensor 2.2, the average error between theoretical value and
measured value is 7.1%. In total five flex sensors are used,
and they are fixed on PIP joints. The flex sensors and the
bending sensors mounted on the glove are firmly attached
to each other, so the bending angles of these two sensors
can be considered to be the same. Data from flex sensor
and bending sensor on the glove are collected at the same
time. Figure 11 and Table 2 show the testing results.
The two lines of the index finger match quite well, and
each cycle perform almost the same. Same tests were
Fig. 10 Testing results of Flex Sensor 2.2. In total nine angles are
tested, and for each angle, ten values are measured. The measured
values closely distribute around the black line, illustrating good
accuracy of the sensor
Fig. 9 Drifting test results. The sensor is kept still in each angle
for 30 s. The resistance drops about 8.5% when the sensor is flat
and drops 3.1 and 1.5% when then bending angle is 60◦ and 120◦
Fig. 11 Testing results of PIP joint of the fabric data-collecting glove.
The two lines match quite well, illustrating excellent accuracy as well
Table 2 Testing results of fabric data-collecting glove
repeated on the other four fingers, and the performance
is as good as the index finger. Variation in all the five
fingers is less than 7%, illustrating the fabric data-collecting
glove has excellent repeatability and precision.
Soft rubber data‑collecting glove
The soft rubber data-collecting glove is made of five
parts, and every part has the same structure with a small
difference in the length of three belts. In this case, we
only exhibit one part of the glove. We choose the part on
the index finger to test, and Flex Sensor 2.2 is also used as
a contrast. The results are shown in Fig. 12.
The results have shown that the two lines match very
well. The visible fluctuations are potentially caused by
unstable connections between conduct wire and the
sensor elongation. The average error between flex
sensor and bending sensor on the soft glove of MP is 7.1%,
illustrating high accuracy of the soft rubber
data-collecting glove. Compared with traditional fabric gloves,
the bionic design of the soft glove makes it possible
for directly sensing the bending motion without any
In the paper, we have presented two different data gloves
based on a same soft bending sensor. The sensor is a
combination of electrical components and mechanical design.
Fig. 12 Testing results of MP joint of the soft rubber data-collecting
glove. The results have proved the excellent repeatability and
precision of the soft sensor glove. And more fluctuation on the red line
shows that the soft rubber data-collecting glove has better sensitivity
than the Flex Sensor 2.2
The sensor has excellent sensitivity as well as
repeatability. Compared with existing bending sensors, the soft
bending sensor is also flexible and can be stretched. The
unrecoverable elongation caused by stretch is avoided by
a novel structure. The size of the sensor can also be
customized. The fabric data-collecting glove solves the
problem of sensor slipping, and the size of the glove can be
fully customized. The soft rubber data-collecting glove,
in which have totally get rid of traditional design
methods, provides a new way for making data gloves. The soft
glove acts like a layer of personal customized skin on the
back of the hand, and bending motion can be measured
directly. The modular design of the glove also simplifies
the design and manufacturing procedures. Besides, both
the sensor and the two kinds of glove have low cost. The
production cost of one such bending sensor is less than 2
dollars, and the costs of two kinds of gloves are less than
30 dollars. Such sensing technology plays a huge role in
promoting data-collecting devices from science
laboratory into clinical use.
In the future, the bending sensor’s repeatability will
be improved by reducing hysteresis using an
appropriate algorithm. The sensor’s sensitivity with respect to
time will be measured. And more bending sensors will be
added on a glove. For example, the rotation of thumb will
be measured to estimate fingertip distance. We also plan
to use this kind of bending sensors to measure the
bending angles of wrist and elbow to capture the motion of the
whole upper extremity.
ZS, JY, MC and ZW helped in the conceptual design; ZS, JY and ZW were
involved in fabrication and testing; XL, ML and YH helped in the design
refinement and electronics; and ZS, ZW, MC were involved in editing the
manuscript. All authors read and approved the final manuscript.
This work was supported by HKU Shenzhen Basic Research Fund
JCYJ20150629151046885, HKU Shenzhen Insititute of Research and
Innovation, and Hong Kong Research Grants Council Early Career Scheme(RGC-ECS)
The authors declare that they have no competing interests.
1. Polygerinos P , Wang Z , Galloway KC , Wood RJ , Walsh CJ . Soft robotic glove for combined assistance and at-home rehabilitation . Robot Auton Syst . 2015 ; 73 : 135 - 43 .
2. Wang Z , Chen MZQ , Yi J. Soft robotics for engineers . HKIE Trans . 2015 ; 22 ( 2 ): 88 - 97 .
3. Polygerinos P , Lyne S , Wang Z , Nicolini LF , Mosadegh B , Whitesides GM , Walsh CJ . Towards a soft pneumatic glove for hand rehabilitation . In: IEEE international conference On intelligent robots and systems. 2013 . p. 1512 - 7 .
4. Overvelde JTB , Mengüç Y , Polygerinos P , Wang Y , Wang Z , Walsh CJ , Wood RJ , Bertoldi K. Mechanical and electrical numerical analysis of soft liquid-embedded deformation sensors analysis . Extreme Mech Lett . 2014 ; 1 : 42 - 6 .
5. Kuang KSC , Cantwell WJ , Scully PJ . An evaluation of a novel plastic optical fibre sensor for axial strain and bend measurements . Meas Sci Technol . 2002 ; 13 ( 10 ): 1523 - 34 .
6. Dunne LE , Walsh P , Smyth B , Caulfield B. Design and evaluation of a wearable optical sensor for monitoring seated spinal posture . In: International symposium on wearable computers (ISWC). 2007 . p. 65 - 8 .
7. Pang F , Liang W , Xiang W , Chen N , Zeng X , Chen Z , Wang T. Temperatureinsensitivity bending sensor based on cladding-mode resonance of special optical fiber . IEEE Photonics Technol Lett . 2009 ; 21 ( 2 ): 76 - 8 .
8. Simone LK , Kamper DG . Design considerations for a wearable monitor to measure finger posture . J Neuroeng Rehabil . 2005 ; 2 : 5 .
9. Gentner R , Classen J. Development and evaluation of a low-cost sensor glove for assessment of human finger movements in neurophysiological settings . J Neurosci Methods . 2009 ; 178 ( 1 ): 138 - 47 .
10. Syed A. Flex sensor based robotic arm controller using micro controller . J Softw Eng Appl . 2012 ; 5 ( 5 ): 364 - 6 .
11. Bakhshi S , Mahoor MH . Development of a wearable sensor system for measuring body joint flexion . In: 2011 International conference on body sensor networks (BSN 2011 ). 2011 . p. 35 - 40 .
12. Saggio G . Mechanical model of flex sensors used to sense finger movements . Sens Actuators A Phys . 2012 ; 185 : 53 - 8 .
13. Tongrod N , Lokavee S , Watthanawisuth N , Tuantranont A , Kerdcharoen T. Design and development of data glove based on printed polymeric sensors and Zigbee networks for human-computer interface . Disabil Rehabil Assist Technol . 2013 ; 8 ( 2 ): 115 - 20 .
14. Kusuda S , Sawano S , Konishi S. Fluid-resistive bending sensor having perfect compatibility with flexible pneumatic balloon actuator . In: 2007 IEEE 20th international conference on micro electro mechanical systems (MEMS). 2007 . p. 615 - 8 .
15. Kramer RK , Majidi C , Sahai R , Wood RJ . Soft curvature sensors for joint angle proprioception . In: IEEE international conference on intelligent robots and systems. 2011 . p. 1919 - 26 .
16. Menguc Y , Park YL , Martinez-Villalpando E , Aubin P , Zisook M , Stirling L , Wood RJ , Walsh CJ . Soft wearable motion sensing suit for lower limb biomechanics measurements . In: IEEE international conference on robotics and automation. 2013 . p. 5309 - 16 .
17. Vogt DM , Park Y-L , Wood RJ . Design and characterization of a soft multiaxis force sensor using embedded microfluidic channels . IEEE Sens J . 2013 ; 13 ( 10 ): 4056 - 64 .
18. Mengüç Y , Park Y L , Pei H , et al. Wearable soft sensing suit for human gait measurement . Int J Robot Res . 2014 . 0278364914543793.
19. Okada O , Keskkula H , Paul D. Mechanical properties of blends of maleated ethylenepropylene rubber and nylon 6 . Polymer . 2001 ; 42 ( 21 ): 8715 - 25 .
20. Hahn P , Krimmer H , Hradetzky A , Lanz U. Quantitative analysis of the linkage between the interphalangeal joints of the index finger. An in vivo study . J Hand Surg . 1995 ; 20 ( 5 ): 696 - 9 .