Content uploaded by Zulkifli Ahmad
Author content
All content in this area was uploaded by Zulkifli Ahmad on Feb 02, 2018
Content may be subject to copyright.
Movement, Health & Exercise, 7(1), 15-25, 2018
15
DEVELOPMENT OF WEARABLE ELECTROMYOGRAM FOR THE
PHYSICAL FATIGUE DETECTION DURING AEROBIC ACTIVITY
Zulkifli Ahmad2*, Mohd Najeb Jamaludin1, and Abdul Hafidz Omar1
1Institute of Human Centered Engineering, Faculty of Biosciences and Medical
Engineering, Universiti Teknologi Malaysia, Malaysia
2Faculty of Mechanical Engineering, Universiti Malaysia Pahang, Malaysia
*Email: zulkola84@gmail.com / kifli@ump.edu.my
(Received 29 December 2017; accepted 8 January 2018; published online 29 January 2018)
To cite this article: Ahmad, Z., Jamaludin, M. N., & Omar, A. H. (2018). Development of
wearable electromyogram for the physical fatigue detection during aerobic activity. Movement,
Health & Exercise, 7(1), 15-25. http://dx.doi.org/10.15282/mohe.v7i1.225
Link to this article: http://dx.doi.org/10.15282/mohe.v7i1.225
Abstract
Physical fatigue or muscle fatigue is a common problem that affects people
who are vigorously involved in activities that require endurance movements.
It becomes more complicated to measure the fatigue level when the dynamic
motion of the activity is included. Therefore, this paper aims to develop a
wearable device that can be used for monitoring physical fatigue condition
during aerobic exercise. A 10-bit analog to digital converter (ADC) micro-
controller board was used to process the data sensed by Ag/AgCl electrodes
and real-time transmitted to the computer through Bluetooth's technology.
The wearable was attached to the knee and connected to the biopotential
electrodes for sensing the muscle movement and convert it into the electrical
signal. The signal then processed by using the fourth-order Butterworth filter
to filter the low-pass filter frequency and eliminate the noise signal. The
results reveal that the fatigue level increased gradually based on the rating of
perceived exertion (RPE), using 10-point Borg's scale, which is rated by the
subject’s feeling. Both muscle's activities in lower limb rise as speed is
increased, and it was also observed that the rectus femoris is functioning
more than gastrocnemius due to the size of muscle fiber. Furthermore, it
was established that the maximum volumetric contraction (MVC) could be
used as a reference and indicator for measuring the percentage of contraction
in pre-fatigue but not to fatigue induced experiment. However, this wearable
device for EMG is promising to measure the muscle signal in the dynamic
motion of movement. Consequently, this device is beneficial for a coach to
monitor their athlete's level of exhaustion to be not over-exercise, which also
can prevent severe injury.
Keywords: Physical fatigue, exercise, wearable device, EMG.
Movement, Health & Exercise, 7(1), 15-25, 2018
16
Introduction
Most sports-race related activities such as marathon, cycling, and long-distance running
require vigorous and repeating movement of the body limbs to win the games. The
determination of the winner of such competitions is based on the endurance capability
of the athletes. This type of sport categorizes as an aerobic activity which depends on the
fitness and competency of the athlete to sustain their energy to the end. Monitoring of the
muscle fatigue during dynamic movement is one of the biggest challenges in wearable
devices. It requires the persistent reliability of the device to detect any type of movement.
In addition, a wearable device to detect physical fatigue solely based on the heart rate,
such as Polar Heart Rate does not represent the muscle fatigue level explicitly.
One of the methods to measure the activity of the musculoskeletal system during sports
activity is by using the surface electromyogram (sEMG) signal. sEMG is a measurement
of muscle response or electrical activity in response to a nerve's stimulation to the muscle
(Elamvazuthi et al., 2015). It is widely used in the clinical (Hawkes et al., 2015),
rehabilitation (Elamvazuthi et al., 2015), ergonomics (Jia & Nussbaum, 2016) and sports
applications (Taha et al., 2017; Wang, Hong, & Li, 2014). Most of the sEMG used in the
previous research are from commercial sensor systems such as Shimmer (Ahmad et al.,
2014; Taha et al., 2017), Step32 (Di Nardo et al., 2016), and Noraxon TeleMyo (Noraxon
USA Inc., Scottsdale, USA) (Sterzing, Frommhold, & Rosenbaum, 2016). Nevertheless,
this sEMG device is possible to be developed to allow for customization of the algorithm
used (Ganesan, Gobee, & Durairajah, 2015). It is interesting to note that such customized
wearable device is capable of detecting and predicting fatigue to up to 90% accuracy (Al-
Mulla, Sepulveda, & Colley, 2011). The decrease of the EMG amplitude signifies that the
muscle is in the fatigue condition for an isometric contraction activity (Ahmad, Najeb,
Amir, & Hafidz, 2017).
There are several existing methods that could be employed to process the sEMG raw data.
Some of the conventional methods used are wavelet analysis, time-frequency approach,
auto-regressive model, and artificial intelligence (Raez et al., 2006; Al-Mulla et al., 2011;
Camic, Kovacs, VanDusseldorp, Hill, & Enquist, 2017; Montgomery, Abt, Dobson, Smith,
& Ditroilo, 2016)). It is worth noting that the aforementioned methods are suitable for
offline measurement of the EMG owing to the ease of data manipulation as well as low-
computing power. Other researchers (Ahmad & Mong, 2016), have also attempted in
online or real-time data processing method via Processing software and immediately plot
the EMG signal graph on the computer. Therefore, the purpose of this study is to develop
a wearable electromyogram device to measure the level of muscle fatigue of an athlete
during aerobic sports activity.
Methodology
Wearable Device Development
Electromyogram (EMG) is an electrical signal that amplifies the physical movement. It
has two-stage amplifiers of the bio-amplifier in between two integrated circuits (IC) op-
Wearable electromyogram for the physical fatigue detection
17
amp with a high-pass filter, which removes any DC generated noise from the electrodes.
For the first stage (refer Figure 1), it amplifies by an instrumented amplifier (INA126),
and the second stage is a standard non-inverting op-amp (OPA347). To sense the muscle
activity, wet electrodes of Ag/AgCl with the 5cm diameter were connected through an
electrode jack. The output signal of the second stage was read by the 10-bit micro-
controller in analog to digital converter (ADC) and powered by 3.3 V operational voltage.
The micro-controller will stack up on the female header of EMG for easy maintenance and
programming processes. The obtained data were transmitted to the computer by HC-05
(Bluetooth module) wirelessly.
Figure 1: Main components circuit (excluded resistor and capacitors)
Since the measurement of EMG was selected on the rectus femoris (thigh) and
gastrocnemius (calf) muscles, therefore, this wearable device was attached near to the knee.
This is to ensure the distance between both electrodes are not far from the main board (as
shown in Figure 2). When the subject was performed the treadmill running, this device can
operate as an independent device by setting the maximum value of contraction for the
buzzer and vibrator to give an alarm to the user. In the data-storage implementation, this
device applicable to send the signal directly to the computer or just keep in the SD card for
offline processing.
Movement, Health & Exercise, 7(1), 15-25, 2018
18
Figure 2: Location of the wearable device on the knee
Subjects
Three male volunteers participated in this study after briefed extensively on the protocol
and method to achieve the objectives of this study. The physical characteristics are
presented in the format of (mean ± SD): Age (31.7 ± 1.2) years, body mass (70.7 ± 4.2)
kg, height (169.3 ± 2.1) cm and BMI (24.6 ± 0.9). These volunteers were from the Sports
Innovation and Technology Group, Faculty of Biosciences and Medical Engineering,
UTM, Malaysia. All of them were completed a health history questionnaire and signed a
written informed consent prior to testing.
Protocol
This protocol was designed for the pre-fatigue, and fatigue induced condition as shown in
Figure 3. Prior to the commencement of the experiment, wet electrodes were attached to
the selected lower limb muscles, i.e., rectus femoris (RF) and gastrocnemius lateralis (GL).
Interelectrode distance is about 5 cm from each location with is targeted on the muscle
belly in order to obtain the maximum active muscle contraction.
Figure 3: Schematic of the experimental protocol.
Wearable electromyogram for the physical fatigue detection
19
Pre-Fatigue
The experiment began by acquiring the required vital sign parameters such as blood
pressure, heart rate, and body temperature were measured in three positions of this protocol
(pre, warm-up and post). This is to investigate the effectiveness of physical activities to
the body’s physiology. Then, isometric of maximum volumetric contraction (MVC) for
Gastrocnemii: subjects were instructed to stand on toe-tips and maximally contract their
shank muscle; while for the Quadriceps: sat on a chair with their hips flexed to 90˚ and
their knee fully extended, they need to resist a force being applied downwards (Ghazwan,
Forrest, Holt, & Whatling, 2017). The procedure was repeated three times for 5 seconds
contraction and 2-second interval. The maximum contraction of the voltage was recorded
as a relative indicator of the fatigue induced signals. After that, the subject was asked to
stand on the treadmill to prepare for 5 minutes walking warm-up with 4 km/h speed. At
the same time, the EMG signals are recorded as well the muscle activities during warm-
up.
Fatigue Induced Protocol
Before the fatigue induced experiment was taken place, the vital sign parameters were
measured again to indicate the effect of warm-up towards the physiology of the body. This
session takes less than 2 minutes to complete with three times measurement. Firstly, the
treadmill speed was set as a warm-up position to make subject used with that acceleration.
After 2 minutes, the speed was increased by 1 km/h. The RPE method is used in the
experiment, and the subject rates their level of fatigue. The protocol will be stopped when
achieved one of the following criteria: 1) exceeds the maximum heart rate; 2) after 30
minutes; and 3) volitional fatigue. The speed increment is maintained and retained at 12
km/h to prevent injury during running due to high-speed movement.
Signal Processing
Raw signals of EMG from the device were acquired to process for analyzing the muscle
contraction during exercise. There are several steps to follow in order to get the clean
signal. Figure 4 depicts the three steps employed by using MATLAB (version R2012a,
Mathworks Inc.), the raw signals were band-pass filtered to remove the movement
artefacts, by a bandwidth 10 – 500 Hz (Hassanlouei, Arendt-Nielsen, Kersting, & Falla,
2012; Montgomery et al., 2016). Prior to that, notch filter 50Hz was used to eliminate
power line noise. Then, it was rectified by the absolute value of the signals into the positive
side, and this method called “full wave rectification." The most important process is low-
pass-filtered by a discrete version of a traditional low-pass filter such as Butterworth or
Chebyshev. The 4th order of Butterworth filter is commonly used by researchers to capture
and “envelope” the signal (Camic et al., 2017; Hsu et al., 2017; Montgomery et al., 2016).
Figure 4: Schematic diagram of the signal processing
Raw Data Filtering Rectifying Low-pass
Filtering
Movement, Health & Exercise, 7(1), 15-25, 2018
20
Results and Discussion
(a) (b)
Figure 5: MVC of muscles for (a) gastrocnemii and (b) quadriceps.
Two types of MVC are applied for the gastrocnemii and quadriceps muscles as indicated
in Figure 5. The function of MVC is to determine the maximum contraction of specific
muscle by different activity provided. Significantly, each MVC will produce the highest
contraction compare with the others as illustrates in Figure
Figure 5(a) and Figure 5(b) illustrate the EMG signals for gastrocnemius and rectus
femoris muscles, respectively. In addition, there are three times of contractions were
performed to obtain the average value of MVC. Mean MVC for gastrocnemii is 7.3843
mV while for quadriceps is 12.4053 mV. There is a significant difference between both
muscles as shown in Figure 5(b), it could be observed that there is no contraction for the
gastrocnemius, whilst the femoris produced the greatest contraction in three attempts. It
demonstrates that the subject performed the proper MVC procedure for the right selected
muscle contraction in Figure 5(b) as compared to what he has done in Figure 5(a).
Figure 6: Muscle activation for gastrocnemius and rectus femoris during warm-up.
Wearable electromyogram for the physical fatigue detection
21
Figure 6 illustrates the muscle activity for 5 minutes warm-up walking on the treadmill
with speed 4km/h. It was observed that the gastrocnemius muscle slightly decreased as the
exercise time increased, and it is contrasted to the rectus femoris muscle. This is due to the
walking pattern that is unique for each individual, and this can be proven via gait analysis.
As compared to the MVC, the maximum contraction for gastrocnemius in pre-fatigue is
near to 6 mV or can be calculated in percent 81.25% MVC whereas for rectus femoris is
approximate to 72.55% MVC. The dynamic force movement in pre-fatigue walking
generates more muscle fiber to be active and produce the pattern for prediction in the
fatigue induced protocol. Since this is a constant load with same walking speed, the
changes of muscle activity slightly differ from the beginning of the exercise.
Figure 7: Comparison between muscles in the fatigue induced protocol.
Based on two selected muscles at the lower limb, Figure 7 indicates the muscle activity
along the fatigue induced experiment. In general, both muscles show the same pattern of
growth as speed increases, however, it is apparent that the rectus femoris is more active
compared to the gastrocnemius. This is might be the size of muscle fiber in the thigh is
bigger than the calf; therefore, the contraction in the quadriceps is stronger than
gastrocnemius. Unfortunately, MVC value in the first place cannot be used in this fatigue
induced protocol since this contraction is higher than MVC. This signifies the MVC
performed in this study is only capable of comparing low-speed movement or static
position exercise. When divided into six regions of stages, naturally the muscles were
much activated at higher speed. It complies with Newton’s 3rd Law of motion, which
considers the action-reaction mechanism of force impact to the ground. This theory is
referring to the ground reaction force (GRF) when the foot touched and contacted to the
ground. The reaction force from GRF works as the normal force exerted from the ground
is equal to the internal force that applied to bone and muscles. Hence, the increasing GRF
is directly reflecting the high muscle activation to make the system to be in an equilibrium
condition. For more details, those regions are extracted as an individual 10-second graph
for gastrocnemius muscle are depicted in Figure 8.
Movement, Health & Exercise, 7(1), 15-25, 2018
22
Figure 8: Specific region for gastrocnemius.
For the first three stages in Figure 8, the EMG fluctuated and gave the mean value 2.3182
mV, 3.9776 mV, and 4.9924 mV. In this stage, the subject was walking instead of running
because the speed in stage III is 6 km/h and still could manage it. Nevertheless, stage IV
and V consistently form the repeated cycle of contraction with an increasing mean value
as 6.6621 mV and 9.8895 mV. Consequently, the transition-to-fatigue are visualized in the
VI stage where the signal was interrupted by the movement artefact caused by fatigue. The
shape of the signal oscillates without a pattern before the end the experiment at 10 minutes
and 21 seconds.
Figure 9: Comparison of gastrocnemius muscle in three subjects
Wearable electromyogram for the physical fatigue detection
23
The final results in this section (Figure 9) explored all the physical activities of each subject
volunteered for this study for the fitness investigations through the fatigue induced
protocol. It is focused on the gastrocnemius muscle alone for the parameter's comparison.
It could be clearly seen that the muscle activity produces the same trend of the signal by
increase steadily over time. From three volunteers, subject 3 had to stop the experiment
earlier due to volitional fatigue while the rest continued to exceed 17 minutes. In this
experiment, body fitness is closely related at the time of aerobic exercise can be sustained,
and it is evident that the subject 1 and 2 has good fitness. On the other hand, muscle activity
for subject 2 is lower compared with the others and might due to the method or position of
the foot during running. Even so, the signal trend of EMG is still acceptable as it rises
steadily until the experiment ends.
Conclusion
As the conclusion, the objective of this study is to develop a wearable device that can be
used in measuring physical fatigue was achieved. This device is capable of monitoring the
whole activity in fatigue protocol started from MVC and ended with volitional fatigue.
The trend and pattern of the signal also indicate the consistency of this wearable device in
determining the EMG.
Acknowledgment
This work was financially supported by the Universiti Teknologi Malaysia under the
Research University Grant, Tier 2 (Q.J130000.2645.13J73) and Ministry of Higher-
Education Malaysia.
References
Ahmad, Z., & Mong, T. C. (2016). Development of Portable Biofeedback Devices for
Sport Applications. In S. I. Ismail, N. Sulaiman, & R. Adnan (Eds.), Proceedings of the
2nd International Colloquium on Sports Science, Exercise, Engineering and
Technology 2015 (ICoSSEET 2015) (pp. 55–66). Singapore: Springer Singapore.
http://doi.org/10.1007/978-981-287-691-1_6
Ahmad, Z., Najeb, M., Amir, M., & Hafidz, A. (2017). Detection of Localised Muscle
Fatigue by Using Wireless Surface Electromyogram (sEMG) and Heart Rate in Sports.
In International Medical Device and Technology Conference (pp. 215–218).
Ahmad, Z., Taha, Z., Hassan, H. A., Hisham, M. A., Johari, N. H., & Kadirgama, K. (2014).
Biomechanics Measurement in Archery. Journal of Mechanical Engineering and
Sciences, 6(1), 762–771. http://doi.org/10.1017/CBO9781107415324.004
Movement, Health & Exercise, 7(1), 15-25, 2018
24
Al-Mulla, M. R., Sepulveda, F., & Colley, M. (2011). An autonomous wearable system
for predicting and detecting localised muscle fatigue. Sensors, 11(2), 1542–1557.
http://doi.org/10.3390/s110201542
Camic, C. L., Kovacs, A. J., VanDusseldorp, T. A., Hill, E. C., & Enquist, E. A. (2017).
Application of the neuromuscular fatigue threshold treadmill test to muscles of the
quadriceps and hamstrings. Journal of Sport and Health Science, (July).
http://doi.org/10.1016/j.jshs.2017.06.002
Di Nardo, F., Mengarelli, A., Burattini, L., Maranesi, E., Agostini, V., Nascimbeni, A., …
Fioretti, S. (2016).
Normative EMG patterns of ankle muscle co-contractions in school-age children during
gait. Gait and Posture, 46, 161–166. http://doi.org/10.1016/j.gaitpost.2016.03.002
Elamvazuthi, I., Zulkifli, Z., Ali, Z., Khan, M. K. A. A., Parasuraman, S., Balaji, M., &
Chandrasekaran, M. (2015). Development of Electromyography Signal Signature for
Forearm Muscle. Procedia Computer Science, 76(Iris), 229–234.
http://doi.org/10.1016/j.procs.2015.12.347
Ganesan, Y., Gobee, S., & Durairajah, V. (2015). Development of an Upper Limb
Exoskeleton for Rehabilitation with Feedback from EMG and IMU Sensor. Procedia
Computer Science, 76(Iris), 53–59. http://doi.org/10.1016/j.procs.2015.12.275
Ghazwan, A., Forrest, S. M., Holt, C. A., & Whatling, G. M. (2017). Can activities of daily
living contribute to EMG normalization for gait analysis? PLoS ONE, 12(4), 1–12.
http://doi.org/10.1371/journal.pone.0174670
Hassanlouei, H., Arendt-Nielsen, L., Kersting, U. G., & Falla, D. (2012). Effect of
exercise-induced fatigue on postural control of the knee. Journal of Electromyography
and Kinesiology, 22(3), 342–347. http://doi.org/10.1016/j.jelekin.2012.01.014
Hawkes, D. H., Alizadehkhaiyat, O., Kemp, G. J., Fisher, A. C., Roebuck, M. M., &
Frostick, S. P. (2015).
Electromyographic assessment of muscle fatigue in massive rotator cuff tear.
Journal of Electromyography and Kinesiology, 25(1), 93–99.
http://doi.org/10.1016/j.jelekin.2014.09.010
Hsu, W.-C., Tseng, L.-W., Chen, F.-C., Wang, L.-J., Yang, W.-W., Lin, Y.-J., & Liu, C.
(2017). Effects of compression garments on surface EMG and physiological responses
during and after distance running. Journal of Sport and Health Science, (January).
http://doi.org/10.1016/j.jshs.2017.01.001
Jia, B., & Nussbaum, M. A. (2016). Development and evaluation of an EMG-based model
to estimate lumbosacral loads during seated work. International Journal of Industrial
Ergonomics, 55, 96–102. http://doi.org/10.1016/j.ergon.2016.08.007
Wearable electromyogram for the physical fatigue detection
25
Montgomery, G., Abt, G., Dobson, C., Smith, T., & Ditroilo, M. (2016). Tibial impacts
and muscle activation during walking, jogging and running when performed
overground, and on motorised and non-motorised treadmills. Gait and Posture, 49,
120–126. http://doi.org/10.1016/j.gaitpost.2016.06.037
Raez, M. B. I., Hussain, M. S., Mohd-Yasin, F., Reaz, M., Hussain, M. S., & Mohd-Yasin,
F. (2006). Techniques of EMG signal analysis: detection, processing, classification and
applications. Biological Procedures Online, 8(1), 11–35. http://doi.org/10.1251/bpo115
Sterzing, T., Frommhold, C., & Rosenbaum, D. (2016). In-shoe plantar pressure
distribution and lower extremity muscle activity patterns of backward compared to
forward running on a treadmill. Gait and Posture, 46, 135–141.
http://doi.org/10.1016/j.gaitpost.2016.03.009
Taha, Z., Rabiu Muazu, M., Mohamad Razali, A., Azrai, M., Razman, M., Chei, M. L., …
Haque, M. (2017). The Application of Inertial Measurement Units and Wearable
Sensors to Measure Selected Physiological Indicators in Archery. Asian Journal of
Pharmaceutical Research and Health Care, 9(2), 85–92.
http://doi.org/10.18311/ajprhc/2017/11046
Wang, L., Hong, Y., & Li, J. X. (2014). Muscular Activity of Lower Extremity Muscles
Running on Treadmill Compared with Different Overground Surfaces. American
Journal of Sports Science and Medicine, 2(4), 161–165. http://doi.org/10.12691/ajssm-
2-4.
