Sensors and Actuators A 274 (2018) 132–140

Contents lists available at ScienceDirect

Sensors and Actuators A: Physical

journal homepage: www.elsevier.com/locate/sna

Human motion monitoring in sports using wearable graphene-coated

fiber sensors
Jinnan Zhang ∗ , Yanghua Cao, Min Qiao, Lingmei Ai, Kaize Sun, Qing Mi, Siyao Zang,
Yong Zuo, Xueguang Yuan, Qi Wang ∗
State Key Laboratory of Information Photonics and Optical Communications, Beijing University of Posts and Telecommunications, Beijing, 100876, People’s
Republic of China

a r t i c l e i n f o a b s t r a c t

Article history: Human motion monitoring is significant for professional athletes who desire to improve their perfor-
Received 8 December 2017 mance in sports. In this study, we have fabricated a wearable graphene-coated fiber sensor and achieved
Received in revised form 10 February 2018 the monitoring of the key motions during playing basketball and soccer. Most of body movements even
Accepted 12 March 2018
the subtle ones, could be accurately identified by the sensor. The recorded sensing data indicated the
Available online 16 March 2018
ultrahigh sensitivity and a very wide sensing range. The high reproducibility and fast response ability
demonstrated during the measurements shows that the as-fabricated sensors have the great potentials
Keywords:
in the applications of sports motion monitoring.
Graphene-coated fiber
Wearable sensor © 2018 Elsevier B.V. All rights reserved.
Human motion
Sports
Sensitivity

1. Introduction the raw data obtained by accelerometers or gyroscopes can not

directly describe the human motions and complex algorithms are
Currently, there is rapidly increasing interest in employing required for the analysis and identification.
wearable, highly flexible sensors in the fields of human health, Strain sensors require relatively simple read-out systems and
motion monitoring and etc. Especially, as sports enthusiasts and simultaneously offer high flexibility and stretchability. Therefore,
professional athletes are continually searching for the opportuni- flexible strain sensors are being widely presented for human
ties to improve their athletic performance and gain a competitive motion monitoring. Trung et al. [18] provided a review of flexible
advantage, more and more wearable sensors have been used for and stretchable physical sensors for applications in activity mon-
strategy, performance enhancement and maintenance, training itoring and personal healthcare. Chen et al. [19] proposed a body
optimization, stress factor estimation, injury prevention, and so on temperature sensor which is wearable, breathable, and stretchable
[1–6]. for healthcare monitoring.
Typically, accelerometers [7–9], gyroscopes [10–12], and flexi- Recently, nanomaterials with excellent nanoscale flexibility and
ble strain sensors [13–16] are used to identify and monitor human electrical properties have frequently been employed in flexible
movements, either individually or in combination. Accelerometers strain sensors, such as carbon nanotubes (CNTs) [20–22], metallic
are highly susceptible to the external disturbances, but the mea- nanowires [23], and graphene [14,15,24–27]. Among those sensors,
sured values are stable enough. A gyroscope can be easily integrated graphene-coated fiber have great potential for the application of
to obtain the angular relations, showing good dynamic perfor- human motion monitoring [26], sustaining tensile strain, bending
mance and little interference from external sources. However, the or twisting to detect the complex movements of the human body. In
measured values of a gyroscope vary greatly with time. The com- addition, their advantage of combining high sensitivity and a broad
bination of the accelerometer and gyroscope seems to be a good sensing range make them suitable for monitoring human actions,
choice, but the high cost will be another problem [17]. Moreover, including pulse, heartbeat, elbow bending, and so on. However, to
the best of our knowledge, there are very few reports on sports
motion monitoring using graphene-coated fiber sensors.
In this work, human motion monitoring in sports, includ-
∗ Corresponding authors.
ing basketball and soccer, was achieved using low-cost wearable
E-mail addresses: [email protected] (J. Zhang), [email protected]
(Q. Wang). strain sensors composed of graphene-coated composite fiber. The

https://doi.org/10.1016/j.sna.2018.03.011
0924-4247/© 2018 Elsevier B.V. All rights reserved.
 J. Zhang et al. / Sensors and Actuators A 274 (2018) 132–140 133

Fig. 2. Electrical resistance of PDCY–RGO after versus numbers of dip-coatings.

Fig. 1. Schematic illustrations of the fabriction process and the characterization of

the graphene-coated fiber. Schematic structure and SEM images of (a) DCY, (b) PDCY,
(c) PDCY–GO (ten dip-coatings), and (d) PDCY–RGO (ten dip-coatings).

measurement and analysis of whole-movement processes were

implemented. The data obtained corresponded to the activities
measured, including accurate and vigorous movement. The results
of the experiments could help the players improve their perfor-
mance, and this new type of fiber sensor will have broad and
practical applications in sports monitoring and many other areas.

2. Fabrication

The scanning electron microscopy (SEM) images of the fiber

sample are shown in Fig. 1a–d. On the left side of each image, Fig. 3. The experimental configuration for strain test of the graphene-coated fiber.

there is a picture of the fiber, and on the right, there is a five-

fold magnification of the same fiber. We used double-covered yarn stable performance. Considering the coating performance and the
(DCY) as the elastic scaffold. This yarn consisted of a highly elas- costs involved, we chose the fiber that had been dip-coated ten
tic polyurethane (PU) core fiber and polyester (PE) fibers winding times for the basical property testing and human motion monitor-
around the PU core helically, just like in [26], as shown in Fig. 1a. The ing in this article, unless otherwise specified.
fibers were etched via air plasma treatment to create a ripple-like
texture on these PE fibers, as shown in Fig. 1b. Due to the generation 3. Strain testing of the fiber
of oxygen-containing polar groups, the yarn’s hydrophilicity was
tremendously improved, after which it was called PDCY (Plasma 3.1. Stretching testing
DCY) [28]. After that, the PDCY was dip-coated with graphene oxide
(GO) (PDCY-GO). The surface of the PE fibers was wrapped in GO The experiment configuration is shown in Fig. 3. To investi-
sheets conformally while the ripple texture was still clearly visi- gate the stretching properties of the fiber, the relative resistance
ble (Fig. 1c). At the end of this process, the PDCY–GO was soaked (R/R0 ) was monitored under various strain conditions by apply-
in hydrobromic (HBr) acid to obtain a reduced GO (RGO)-coated ing a constant current across the fiber, as shown in Fig. 3. The copper
fiber (PDCY–RGO, Fig. 1d). By repeating the process, multiple coat- wires were glued to both ends of the fiber by conductive copper
ings could be obtained. The fabrication details can be found in our foil tapes and then connected to the probers of a constant current
previous reports [29,30]. source meter (Keithley Series 2600 A). The current applied to the
The PDCY–RGO fiber was light (only about 0.28 g/m) and flexible fiber was set at 5 nA. The voltage values of the fiber under various
enough to conform to any position of the human body, including strains were uploaded to a computer via a serial port. The data were
those of the throat, finger, knee, and so on. No matter how many recorded by a program we developed.
times the fibers were dip-coated, we found that the resistance was The fiber was fixed on a computer-controlled, motored linear
linearly proportional to the length of fibers when the fibers were stage to test its response to stretching. The stretching was applied
stretched axially. This indicates uniform GO coating during the dip- along the length of the graphene fiber, as shown in Fig. 4a. The
coating process. stretching and releasing processes of the fiber were first recorded
The resistance of the composite fiber varies with the number up to 80% strain. Fig. 4b depicts the resistance variation of both pro-
of dip-coatings. The relationship between the unit resistance and cesses. The resistance increased monotonically with tensile strain
number of dip-coatings is shown in Fig. 2: the resistance decreased up to 80%. It is worth noting that up to 50% strain, the change in
from 7.72 M /m to 0.71M/m as the number of dip-coatings resistance was almost proportional to the change in strain within
increased from two to twenty. When the number of dip-coatings this range. Relatively good linearity was observed. The graphene-
was less than ten, the unit resistance changed obviously. However, coated fiber maintained stable and accurate performance under
the unit resistance changed only gradually when the number of dip- both low and high strains.
coatings was more than ten. The mechanical properties of the fiber Durability and stability are critical for practical application. The
were also evaluated. The rupture strain of the fiber reached up to cycling test was performed under a strain of 50%. As shown in
229%. The electrical resistance of graphene-coated fibers indicated Fig. 5a, the electrical response exhibited high stability over 2000
134 J. Zhang et al. / Sensors and Actuators A 274 (2018) 132–140

Fig. 6. Resistance variation in the bending fiber during the bending and releasing
process.

Fig. 4. a) Electro-mechanical universal testing machines for fiber stretching, b)

Resistance variation of the fiber under a stretching-releasing cycle (80% strain).

Fig. 7. The cycling test of the bending.

the fiber under various bending angles. The results shown in Fig. 6
indicate that we could detect more than 150◦ of bend change using
Fig. 5. a) The cycling test of the fiber under a strain of 50%, b) A zoomed-in view of
the fiber. These features make it possible to detect the activities
the cycling test. of large human joints, such as the knee and elbow. A hysteresis is
observed. So there is a few differences between the performances
of the bending and releasing process.The hysteresis is very low and
cycles (stretching and releasing). The maximum relative resistance the original resistance of the fiber sensor can be finally recovered
of the fiber was stable at about 1.2. We selected five cycles (from after releasing it from strain. We think the hysteresis observed is
1601 to 1605) from the 2000 cycles for demonstration purposes, mainly related to the friction between the PE fiber winding and
as shown in Fig. 5b. From the zoomed-in view of the cycle, we can silicone gel substrate. The friction hinders the fiber from timely
observe periodic changes in relative resistance. Waveform changes recovery to the unstrained state after the substrate releasing, thus
were not obvious during the five cycles. This indicated that the inducing the hysteresis of the fiber itself [16].
performance of the fiber was stable. The cycling test of bending was performed from 0◦ to 150◦ . The
curves of the various bending times (from 1 to 2000) are shown in
3.2. Bending testing Fig. 7. There was little difference between the curves, so the fiber
showed high bending stability. Compared to the stretching test, the
The bending ability of the fiber is of great concern for human relative resistance changed more smoothly due to the silicone gel
motion monitoring in sports. We carved a groove on the silicone substrate.
gel substrate to fix the fiber (inset in Fig. 6). Thus, the fibers could
be bent together with the silicone gel substrate. The bending angle 3.3. Torsion testing
␪ is defined in the inset of Fig. 6. The fiber was also connected to
the probers of the costant current source meter by copper wires. In order to implement the torsion test, we used a bracket to
The two ends of the silicone gel were pressed to cause it to bend, fix the fiber sensor, as shown in Fig. 8a. One end was fixed at the
which caused the bending and stretching of the fibers, resulting bottom of the bracket, while the other end was fixed at the top and
in resistance changes. In this way, we could test the changes in could be rotated. The bracket was marked with a scale to indicate
 J. Zhang et al. / Sensors and Actuators A 274 (2018) 132–140 135

Fig. 9. The cycling test for torsion.

Fig. 8. a) Testing configuration for torsional strain, b) Resistance variation of the

graphene-coated fiber during the torsion test. (For interpretation of the references
to colour in text, the reader is referred to the web version of this article).

the amount of torsion. We first twisted along the fiber-winding

direction (clockwise) from 0 to 12␲ rad. Then, we twisted the fiber
in the negative direction (opposite to the fiber-winding direction,
counterclockwise) from 0 to −12␲ rad.
The torsion results are shown in Fig. 8b. The R/R0 versus tor-
sion level curve demonstrated an outstanding wide dynamic range
for torsion sensing(from −12␲ to +12␲ using 3.5 cm yarn), one
superior to that of conventional torsion sensor [27]. Moreover, Fig. 10. A schematic diagram showing human motions, including subtle and large
motions, as well as the fiber sensors on each joint.
because the fiber itself had a particular winding direction, there was
a difference in the tendency of resistance changes for the clockwise
and counterclockwise rotations. 4. Human motion monitoring in sports
When the twist direction was consistent with the fiber-winding
direction, the fiber’s outer layer became more compact, and the 4.1. Monitoring of human montions
contact area became larger, leading to a decrease in relative resis-
tance, and vice versa. As shown in Fig. 8b, the relative resistance In order to monitor human motions, we must make fibers into
was reduced to −34.27% under the clockwise rotation (red line) wearable sensors that can be placed on the human body. In this
from 0 rad to 12␲ rad, and it was increased to +65.33% under the research, we fabricated a wearable sensor with a graphene-coated
counterclockwise rotation (black line) from 0 to −12␲ rad. This fiber. The fiber was assembled on an ultrathin silicone gel substrate,
property represents an advantage in distinguishing between dif- and the copper wires used for connections were wrapped with insu-
ferent twist directions. The torsion test was repeated for more than lating tape. Thus, a simple wearable graphene-coated fiber sensor
2000 times as shown in Fig. 9. Over these 2000 repetitions, the rel- was made. The wearable sensors were flexible and light enough
ative resistance at the same degree of rotation underwent only a that they could be stuck to the body joint using insulating tape.
small change. Thus, stability and reproducibility have been proven In this way, we could carry out the full-range detection of human
once more. motions with the wearable graphene-coated fiber sensor.
Stretching, bending, and torsion are very common in human We attached the wearable fiber sensors to several joints of the
movements. In a variety of sports, many key motions include these human body, including the shoulder, elbow, wrist, knee, and ankle,
three strains, or a combination of these three strains [16]. The to monitor the basic movements of the human body and the com-
graphene-coated fiber not only has multiple functions in terms of plex movements involved in specific sports, as shown in Fig. 10.
stretching, bending, and torsion but also shows excellent perfor- The activities of the elbow are typical of major motions. On the
mance in terms of a combination of high sensitivity and a broad other hand, wrist activities are typical of the precise motions of the
sensing range. These features allow its application as a full-range human body. Therefore, we chose the motions of these two joints
human motion and activity sensor. In addition, the fiber showed to perform basic human movement tests.
stable performance in all tests. The results clearly show that the Firstly, we tested elbow flexion and extension. The tester sat at a
sensor has potential for practical applications in human motion table with his arms stretched out on that table. The fiber was fixed
detection. on the elbow. When the test began, the tester bent his arm, held it at
136 J. Zhang et al. / Sensors and Actuators A 274 (2018) 132–140

Fig. 11. The signals of elbow. Fig. 13. The signals of wrist rotation.

Fig. 14. Flow of jump shot motions.

The wrist pulse is also a key physiological signal for human

Fig. 12. The signals of wrist bend. health. We monitored wrist pulse with the wearable graphene-
coated fiber sensor. The data from the sensor corresponded to the
systolic and diastolic blood pressures. These results are published
the maximum bend for 1–2 seconds, and then, straightened it out. in another article [30] and will not be described detailedly here.
This process was repeated ten times. Part of sample data obtained
from the sensor is shown in Fig. 11.
From the data waveform, all the changes in and maintenance 4.2. Monitoring of basketball jump shot
of the basic motions can be observed. The relative resistance rose
when the elbow was bent. While the arm was kept bent, the rela- The monitoring of sports activities is a typical usage scenario
tive resistance was maintained above 150%. When the elbow was for wearable sensors [9,34]. We chose the two most popular sports
straightened, the resistance dropped back to 0%. An overshoot could worldwide, basketball and soccer, for test scenarios.
be observed at the instant of the action switch. In fact, when the We first monitored the basketball jump shot using the wearable
strain was suddenly reversed at the end of the stretching cycle, the graphene-coated fiber sensor. In order to increase the probability
polymers attempted to release their stress instantly via mechani- of a shot going into the basket, a player must develop good shooting
cal deformations [31–33]. In the process of flexion and extension, form.
tiny muscles contracted and shook. As a result, many small sig- The motions of a player shooting a basketball into the basket can
nal changes could be observed. However, these could be easily be resolved into four phases: The ‘catch,’ when the player catches
removed with a low-pass filter. the ball; the ‘buildup,’ when the player raises the ball in front of
The wrist is a very flexible joint that undergoes various bend- the forehead; the ‘release,’ when the player jumps up and then ball
ing and rotation movements. The test procedure is similar to the is released from the hand; and the ‘return,’ which is the recovery
elbow test. The tester sat at the table, relaxed his arms and wrists, process that allows a new cycle to begin after the release, as shown
and placed them on the table. The fiber was fixed on the wrist. in Fig. 14.
When the test began, the tester bent (rotated) his wrist, held it at Several wearable sensors were directly fixed on a basketball
the maximum bend (rotating) for 1–2 seconds, and then restored player’s joints, including the wrist, elbow, and shoulder, to mon-
it to the original position. This process was repeated ten times. itor his shooting motions, as shown in Fig. 10. In our experiment,
Parts of the wrist bending and rotation data sample are shown in the player performed continuous jump shots until ten jump shots
Figs. 12 and 13, respectively. were completed.
Although the wrist’s motion range was far less than that of the The data from the sensors on the wrist, shoulder and elbow were
elbow, relative resistance changes were still obvious and corre- recorded. We chose the raw data obtained from six jumpers as a
sponded to the motion. The difference is that the proportion of demonstration, as depicted in Fig. 15. The data from all three joints
overshoots was relatively larger due to smaller movements of the showed good periodicity, although the three joints began at dif-
wrist. In addition, the motion of the wrist rotation was relatively ferent times during the jump shot. At the same time, the signals of
complex and even included flexion and extension, so the signal was the three joints did correspond with one another. This indicates the
different due to this flexion and extension. stable performance of our sensors.
 J. Zhang et al. / Sensors and Actuators A 274 (2018) 132–140 137

Fig. 15. Response curves of wearable sensors on the wrist, shoulder, and elbow.

Fig. 17. The motion of a) wrist pressing, b) no wrist pressing, and (c) the signals of
the wearable sensor on the wrist.

resistance of the wrist sensor dropped to 0% after little undulation,

the relative resistance of the shoulder sensor dropped from 170%
to 20%, and the relative resistance of the elbow rose to about 170%
after undulation. During the ‘release’ phase, from C to D, the player
holds the shoulder still and pushes the ball with his wrist and elbow.
We can find the quick movements of the wrist and elbow in the
curve.As shown in the fourth image from the left in Fig. 14, the arm
position of the player looks a little bit higher due to the fact the
body is tilted backwards (well-known as the fadeaway jump shot).
In fact, the player does not move the shoulder and arm again (the
angle between the arm and shoulder is almost unchanged, resulting
in the nearly constant relative resistance of the shoulder sensor).In
addition, the jitter of the shoulder sensor caused by the final push-
ing ball is also accurately recorded by the resistance fluctuation
Fig. 16. The signals from the wearable sensors on the wrist, shoulder, and elbow for from time C to D.During the ‘return’ phase after shooting, from D to
entire jump shot process. E, the player’s wrist, elbow, and shoulder return to a relaxed state.
As we can see from the curve in Fig. 16, the relative resistance of the
In order to understand the details of the jump shot and the wrist sensor dropped to 50%, the relative resistance of the shoul-
specific performance of the sensor, we selected the first cycle for der sensor rose to 170%, and the relative resistance of the elbow
detailed analysis, as shown in Fig. 16. We recorded the timing of dropped to 0% after undulation. All the details of the jump shot
each important step (A, B, C, D and E in Fig. 14) in the jump shot, motions were reflected in the data curve.
and marked it on the data for the first cycle. Thus, the time from A In a jump shot, there is one detail that affects both shot accu-
to E in Fig. 16 corresponds to the actual motion. racy and range. This is the motion of wrist pressing (Fig. 17a). Most
In the ‘catch’ phase, the player holds the ball, while his shoulder people ignore this wrist motion and not press the wrist (Fig. 17b).
and arm are kept still. When the phases switch (time B), the player’s Thus, they do not understand the feeling of wrist pressure. At the
elbow bends and prepares to lift the ball to another phase. Thus, the end of the shooting, the player’s hand maintaining this wrist press-
relative resistance of the elbow sensor exhibits a small fluctuation ing shows that the shooting process is steady and consistent. Thus,
near time B. During the ‘build up’ from B to C, the player holds the it could improve the success rate of shooting.
ball. His wrist bends backwards and prepares to shoot, his arm is We asked the player perform shooting training by adopted dif-
lifted, and his elbow is open. His arm is raised to the highest posi- ferent motions (with and without wrist pressing) and complete ten
tion via the shoulder joint motion. The fiber sensor on his shoulder repetitions for each motion. The success rates were 80% with wrist
changes from its maximum length to the minimum one.The relative pressing and 60% with no wrist pressing. Compared to the correct
138 J. Zhang et al. / Sensors and Actuators A 274 (2018) 132–140

Fig. 18. Flow of soccer kicking motions.

Fig. 20. The signals of the wearable sensors on the knee, ankle, and elbow for the
entire kicking process.

the signals of the three joints corresponded to one another at a

given moment. The stable performance of our sensor was verified
once again.
Fig. 19. Response curves of the wearable sensor on the knee, ankle, and elbow. A detailed analysis of the shooting motion is also necessary in
soccer. We chose data from the first complete shooting process for
analysis, as shown in Fig. 20. The time from A to E in Fig. 20 corre-
movements (with wrist pressing), the relative resistance showed sponds to this motion. In the ‘ready’ phase, from A to B, the player’s
less of an increase without wrist pressing from C to D, during the knees and ankles slightly tightened as compared to standing state,
‘release’ phase, as shown in Fig. 17c. During the ‘return’ phase, from while his elbow hardly moved. We can observe that the relative
D to E, obvious resistance changes were observed due to the recov- resistances of the sensors on the knee and ankle exhibited slight
ery of wrist pressing, while resistance hardly changed without wrist fluctuations. In the ‘building up’ phase, from B to C, the large swing
pressing. Based on these signals, we can provide feedback to play- of the leg was evident through the changing resistance curves of the
ers to improve their shooting success rates. Compared with other knee and ankle. Even tiny muscle shaking could be observed. In con-
reports [17,35–37], using sensors, we can not only recognize the trast, the elbow hardly changed at this time. During the ‘hit’ phase,
actions taken by players on the field but also determine the quality from C to D, the resistance on the knee decreased, reflecting the
of these movements based on the details of the motions. knee’s change from closed to open. The player’s ankle movements
were similar. In order to maintain balance, the elbow tightened to
4.3. Monitoring of soccer shooting and passing the chest. Thus, the resistance increased. In addition, the feet and
ankle movements changed very quickly when shooting (<100 ms).
We also monitored the shooting motions in soccer. Two wear- The recorded data indicated a quick response on the part of the
able fiber sensors were attached to the front of a kicking leg’s ankle sensor. By observing the data curves of the sensors placed on the
and knee. Another fiber sensor was applied at the elbow to record knee, ankle and elbow, we can identify all the motions.
the swing of the arm, as shown in Fig. 10. In addition to shooting, passing is also a very important skill in
A entire process of a soccer shot is divided into four phases: the soccer. A strong shot typically involves an instep kick (Fig. 21a). A
‘ready’, when the player gets ready; the ‘buildup’, when the player short pass (short-distance pass) may involve an arch push (Fig. 21b),
is ready to kick in front of the ball; the ‘hit’, when the player kicks which can ensure the highest level of accuracy and safety and the
the ball; and the ‘return’, the recover process that allows a new lowest level of difficulty. A long pass (long-distance pass) may
round of the cycle to begin, as shown in Fig. 18. involve a conventional foot instep kick (Fig. 21c). Players can also
We asked the player to complete seven consecutive shooting produce changes by controlling the toes.
attempts. The corresponding data were recorded, as shown in The three technical motions were monitored using the fiber sen-
Fig. 19. Because of the different ranges of motion of the various sor, and the signal curves are shown in Fig. 22. In the ‘ready’ phase,
joints, the range of relative resistance also varied. Similar to the from A to B, in the curve, the signal of the pass was similar to that
basketball test, the data for all three joints show good periodicity, of a shot. In the ‘building up’ phase, from B to C, the relative resis-
although the three joints start moving at different times. However, tance varied slightly between long pass and the shot due to their
 J. Zhang et al. / Sensors and Actuators A 274 (2018) 132–140 139

5. Conclusion

We have achieved human motion monitoring for sports using a

wearable graphene-coated fiber sensor. In particular, key motions
in playing basketball and soccer have been monitored successfully,
benefiting from the fiber virtue of the high sensitivity, large sens-
ing range and the sensing ability to multiple forms of deformation.
Through processing and analyzing the raw sensing data, various
joint motions can be identified and recorded, which will greatly
Fig. 21. Different ankle motions for (a) shooting, (b) a short pass, and (c) a long pass
help the athletes enhance their performance and optimize their
in soccor. training.Our results clearly show that wearable graphene-coated
fiber sensors which has been fabricated in a low-cost and very easy
way can not only be successfully used in human motion monitoring
in sports but also have very broad application prospects in the field
of monitoring human health.

Acknowledgments

This work was supported by the National Natural Science Foun-

dation of China (NSFC) (Nos. 61474008 and 61571067), Program for
New Century Excellent Talents in University of China (No. NCET-
13-0686) and the Fund of the State Key Laboratory of Information
Photonics and Optical Communications (Beijing University of Posts
and Telecommunications) (No. BUPT600100195).

References

[1] M. Cornacchia, K. Ozcan, Y. Zheng, IEEE Sens. J. 17 (2) (2017) 386–403.

[2] M. Cardinale, M.C. Varley, Int. J. Sports Physiol. Perform. 12 (S2) (2017) 55–62.
[3] M. Kos, I. Kramberger, IEEE Access 5 (2017) 6411–6420.
[4] G. Acampora, D.J. Cook, P. Rashidi, Proc. IEEE 101 (12) (2013) 2470–2494.
[5] M. Chen, S. Gonzalez, A. Vasilakos, H. Cao, V.C.M. Leung, Mobile Networks
Appl. 16 (2) (2011) 171–193.
[6] G. Fortino, G.D. Fatta, M. Pathan, A.V. Vasilakos, Wirel. Networks 20 (7) (2014)
1925–1938.
[7] N.B. Nordsborg, H.G. Espinosa, D.V. Thiel, Procedia Eng. 72 (2014) 132–137.
[8] W. Kim, M. Kim, Soccer kick detection using a wearable sensor, in:
International Conference on Information and Communication Technology
Convergence (ICTC), Jeju, South Korea, 19-21 Oct., 2016.
[9] S.R. Vanrell, D.H. Milone, H.L. Rufiner, Assessment of homomorphic analysis
for human activity recognition from acceleration signals, IEEE J. Biomed.
Health Inf. PP (99) (2017), http://dx.doi.org/10.1109/JBHI.2017.2722870,
pages: 1–1.
[10] J.-K. Min, S.-B. Cho, Activity recognition based on wearable sensors using
selection/fusion hybrid ensemble, in: IEEE Int. Conf. Syst., Man, Cybern.
(SMC), Anchorage, AK, USA, 9-12 Oct., 2011.
Fig. 22. The signals from the wearable sensor on the ankle in a shot, a long pass and
[11] A. Ahmadi, D. Rowlands, D.A. James, Sports Technol. 2 (3–4) (2009) 129–136.
a short pass. [12] E.M. Kidman, M.J.A. D’Souza, S.P.N. Singh, A wearable device with inertial
motion tracking and vibro-tactile feedback for aesthetic sport athletes diving
coach monitor, in: 10th International Conference on Signal Processing and
Communication Systems (ICSPCS), Gold Coast, QLD, Australia, 19-21 Dec.,
different powers. However, because the force was small for the 2016.
short pass, the resistance changes were also small. During the ‘hit’ [13] M. Amjadi, A. Pichitpajongkit, S. Lee, et al., ACS Nano 8 (5) (2014) 5154–5163.
phase, from C to D, the player stood on tiptoe to achieve greater [14] Y. Wang, R. Yang, Z. Shi, et al., ACS Nano 5 (5) (2011) 3645–3650.
[15] H. Huang, S. Ke, B. Wang, H. Long, K. Wang, P. Lu, J. Lightwave Technol. 35 (2)
pass height, which resulted in a resistance drop for the long pass. (2017) 320–324.
However, unlike the long pass, the resistance drop took place at [16] M. Amjadi, K.U. Kyung, I. Park, M. Sitti, Adv. Funct. Mater. 26 (11) (2016)
the next stage during passing. This was because the instep remains 1678–1698.
[17] W. Kajiro, A. Taniguchi, K. Watanabe, Y. Kurihara, Measurement and analyze
tight until the recovery stage. For the short pass, we only observed of jump shoot motion in basketball using a 3-D acceleration and gyroscopic
a slight change in resistance because the ankle motions are funda- sensor, in: 2012 SICE Annual Conference, Akita University, Akita, Japan,
mentally the same from A to E. Note that in soccer, ankle motions August 20-23, 2012.
[18] T.Q. Trung, S. Ramasundaram, B.-U. Hwang, N.-E. Lee, Adv. Mater. 28 (3)
are very subtle and therefore difficult to detect. Nevertheless, we (2016) 502–509.
were still able to monitor the changes of the ankle joint and iden- [19] Y. Chen, B. Lu, Y. Chen, X. Feng, Sci. Rep. 5 (2015) 11505.
tify the corresponding technical actions. Compared with the recent [20] S. Yao, X. Nie, X. Yu, et al., IEEE Sens. Lett. 1 (3) (2016) 2500404.
[21] S. Luo, T. Liu, Adv. Mater. 25 (39) (2013) 5650–5657.
reports [8,38,39], we achieved the more accurate identification of
[22] S.-H. Hwang, H.W. Park, Y.-B. Park, M.-K. Um, J.-H. Byun, S. Kwon, Sci. Technol.
motions using wearable graphene-coated fiber sensors. Our test 89 (2013) 1–9.
scheme was low-cost and easy to use. [23] S. Yao, Y. Zhu, Nanoscale 6 (2014) 2345–2352.
We captured the typical action of soccer and basketball. Apply- [24] S.-H. Bae, Y. Lee, B.K. Sharma, H.-J. Lee, J.-H. Kim, J.-H. Ahn, Carbon 51 (2013)
236–242.
ing the fiber sensor, we achieved the accurate monitoring of the [25] X.-W. Fu, Z.-M. Liao, J.-X. Zhou, et al., Appl. Phys. Lett. 99 (2011) 213107.
motions of the athlete’s multiple joints. Furthermore, based on the [26] Y. Cheng, R. Wang, J. Sun, L. Gao, Adv. Mater. 27 (45) (2015) 7365–7371.
raw data obtained from the sensors in various positions, the motion [27] T. Yang, Y. Wang, X. Li, et al., Nanoscale 6 (2014) 13053–13059.
[28] L. Yang, J. Chen, Y. Guo, Z. Zhang, Appl. Surf. Sci. 255 (2009) 4446–4451.
types could be accurately identified, and the athlete’s motions could [29] Q. Mi, Q. Wang, S. Zang, G. Mao, J. Zhang, X. Ren, Smart Mater. Struct. 27
be improved. (2018) 015014.
140 J. Zhang et al. / Sensors and Actuators A 274 (2018) 132–140

[30] S. Zang, Q. Wang, Q. Mi, J. Zhang, X. Ren, Sens. Actuators A 267 (2017) at Beijing University of Posts and Telecommunications (IPOC, BUPT). Her research
532–537. interests include human health monitoring and management and nanometer mate-
[31] A.D. Smith, F. Niklaus, A. Paussa, et al., Nano Lett. 13 (7) (2013) 3237–3242. rial optical fiber sensing.
[32] Md Omar, Faruk Emon, Jae-Won Choi, Sensors 17 (3) (2017) 656.
[33] H. Liu, M. Dong, W. Huang, et al., J. Mater. Chem. C 5 (2017) 73–83. Lingmei Ai received the bachelor’s degree in electronic information engineering
[34] M. Sharma, R. Srivastava, A. Anand, Wearable motion sensor based phasic from Qingdao University of Technology, China. She is currently a graduate student
analysis of tennis serve for performance feedback, in: IEEE International majoring in electronic science and technology at the State Key Laboratory of Infor-
Conference on Acoustics, Speech and Signal Processing (ICASSP), New mation Photonics and Optical Communications at Beijing University of Posts and
Orleans, LA, USA, 5-9 March, 2017. Telecommunications (IPOC, BUPT). Her research interests include fiber materials
[35] L. Bai, C. Efstratiou, C.S. Ang, weSport: Utilising wrist-band sensing to detect and DSP.
player activities in basketball games, in: IEEE International Conference on
Kaize Sun received the bachelor’s degree in Electronics Science and Technology
Pervasive Computing and Communication Workshops, Sydney, NSW,
from Central South University of Forestry and Technology (CSUFT), China. He is cur-
Australia, 14–19 March, 2016.
rently studying for a master’s degree in the State Key Laboratory of Information
[36] F. Chen, C. De Vleeschouwer, Automatic production of personalized basketball
Photonics and Optical Communications at Beijing University of Posts and Telecom-
video summaries from multi-sensored data, in: 17th IEEE International
munications (IPOC, BUPT). His research interests include human health monitoring
Conference on Image Processing (ICIP), Hong Kong, China, 26-29 Sept., 2010.
and management and nanometer material optical fiber sensing.
[37] Chih-Hsien Hsia, Chin-Hsiang Chien, Hao-Wei Hsu, et al., Analyses of
basketball player field goal shooting pos-tures for player motion correction Qing Mi received B.E. degree from the Department of Materials Science and Tech-
using kinect sensor, in: International Symposium on Intelligent Signal nology, Hebei University of Technology, China in 2015. She is now pursuing an M.E.
Processing and Communication Systems (ISPACS), Kuching, Malaysia, 1-4 degree at Beijing University of Posts and Telecommunications. Her research interest
Dec., 2014. includes graphene, wearable sensors, and multifunctional sensing.
[38] Y. Weizman, F.K. Fuss, Procedia Eng. 112 (2015) 157–162.
[39] H.M.Sajjad Hossain, Md Abdullah Al Hafiz Khan, Nirmalya Roy, SoccerMate: A Siyao Zang received B.E. degree from the Department of Communication Engineer-
Personal Soccer Attribute Profiler using Wearables, in: IEEE International ing, Jilin University, China in 2016. She is now pursuing a Ph.D. at the State Key
Conference on Pervasive Computing and Communications Workshops Laboratory of Information Photonics and Optical Communications (IPOC) at Bei-
(PerCom Workshops), Kona, HI, USA, 13-17 March, 2017. jing University of Posts and Telecommunications. Her research interest includes
graphene, nanometer material sensing, wearable sensors, and vital signs monitor-
ing.
Biographies
Yong Zuo is an Associate Professor at the State Key Laboratory of Information
Photonics and Optical Communications (IPOC), Beijing University of Posts and
Jinnan Zhang received a Ph.D. degree in electromagnetic and microwave technol- Telecommunications (BUPT), China. He received his Ph.D. degree in electronic engi-
ogy from Beijing University of Posts and Telecommunications (BUPT), China. He neering from Queen Mary University of London (QMUL), UK. His current research
is currently a lecture with the State Key Laboratory of Information Photonics and interest includes graphene, nanometer material sensing and wearable sensors.
Optical Communications (IPOC, BUPT) for intelligent sensing. His research interests
include human health monitoring and management, nanometer material optical Xueguang Yuan received the Ph.D. degree in electromagnetic and microwave tech-
fiber sensing, and DSP. nology from Beijing University of Posts and Telecommunications (BUPT), China. He is
currently a lecturer with the State Key Laboratory of Information Photonics and Opti-
Yanghua Cao received a bachelor’s degree in information engineering from Bei- cal Communications (IPOC, BUPT). His research interests include health monitor,
jing University of Posts and Telecommunications (BUPT), China. He is currently a optical sensing, and DSP.
postgraduate student pursuing a master’s degree in intelligent sensing with the
State Key Laboratory of Information Photonics and Optical Communications (IPOC, Qi Wang received a Ph.D. degree from Beijing University of Posts and Telecommu-
BUPT). His research interests include nanometer material optical fiber sensing, social nications, China, in 2005. He is currently a Professor in the Institute of Information
information construction and management, and DSP. Photonics and Optical Communications. He is the author of more than 100 publica-
tions in international journals and conference proceedings. His research interests
Min Qiao received the bachelor’s degree in computer science and technology from include wearable sensors, two-dimensional materials, self-rolled-up micro/nano
Jinan University (JNU), China. She is currently studying for a master’s degree in technology, and III-V semiconductor photoelectronic devices.
the State Key Laboratory of Information Photonics and Optical Communications