2021 8th NAFOSTED Conference on Information and Computer Science (NICS) Wireless Wearable ElectroMyography Acquisition System Utilizing Reduced-Graphene-Oxide Based Sensor Loan Pham-Nguyen1 , Huy-Dung Han1 , Le-Lan Tran1,3 , Nhung Nguyen-Hong1 , Linh T Le2,3 Hanoi University of Science and Technology, Hanoi, Vietnam Flextrapower, Inc., Long Island City, New York, 11101, USA Bonbouton Incorporation, Ho Chi Minh City, Vietnam Abstract—In this work, a complete wireless wearable ElectroMyography (EMG) acquisition and analysis system, which employs reduced-graphene-oxide (rGO) coated-Nylon/Polyesterfabric sensors, is proposed The electronic acquisition system is confirmed by comparing its measured EMG signal to the ones acquired by the commercial circuit Myoware The measurements are carried out on bicep muscle and calf muscle The results suggested that graphene-based smart fabrics can be used as a viable alternative to non-reusable Ag/AgCl electrodes for highquality EMG monitoring The proposed EMG data acquisition and analysis device is small and light-weight providing a smartclothes platform for convenient health tracking Index Terms—ElectroMyography, e-Textiles, health monitoring, reduced graphene oxide I I NTRODUCTION Nowadays, the demand of wearable biomedical devices for health monitoring increases significantly due to its low-cost and feasibility, that may reduce hospital visits for patients Different biomedical signals (Electromyography (EMG), electrocardiography (ECG), skin temperature etc.) related to human body can bring out different useful information on health condition EMG is an myoelectric signal that is formed by physiological variations and expressed through the muscle fiber membranes EMG signals allow to evaluate prosthetic applications [1], recovering program for muscular skeletal disorders [2], and muscle monitoring in sports [3] To acquire EMG signals, certain types of EMG sensors are proposed and investigated Silver/Silver chloride (Ag/AgCl) is a commercial EMG sensor and favored for its non-invasive and easy-handling characteristic However, the wet-gel to stick the sensor to the human skin causes the discomfort that prohibits its usage in continuous monitoring applications A variety of dry electrodes derived from metals are also attractive and studied over the past decade [4], [5], [6] but yet to deliver expected performance These sensors are usually rigid and still cause inconvenience for users Recently, textile based sensors draw more attention for certain advantages comparing to the commercial Ag/AgCl and the dry ones Gel is not required to have contact between human skin and textile-based sensor that remove completely all the negative feelings Besides, it is feasible to incorporate textile-based sensor to clothes 978-1-6654-1001-4/21/$31.00 ©2021 IEEE which is considered as the most suitable sensor for wearable applications Moreover, the physiological signals detected by textile based sensor is comparable with the commercialized Ag/AgCl-based ones [7], [8] An EMG acquisition circuit employed for wearable biomedical applications should be reliable, lightweight and low cost Commercial myoware developed by Advancer Technology Company is quite popular and being employed in some recent EMG acquisition system However, this device shows some disadvantages such as certain noise is not effectively rejected resulting in a low SNR [9], and the wireless communication is not yet integrated In this work, a fully wearable EMG monitoring and analysis system using reduced-graphene-oxide (rGO) coated Nylon/Polyester-fabric sensor is proposed The performance of the system applying graphene-based sensor are compared with the one using disposable Ag/AgCl sensor The experiments on biceps muscles and calf muscles show the versatility of the designed system II S YSTEM DESIGN The block diagram of proposed EMG acquisition system is shown in Figure This EMG acquisition system consists of EMG electrodes, analog acquisition circuit, wireless communication module, and a software written for Android smartphone considered as data-display device To acquire EMG signal from a certain muscle bundle, a system of three identical EMG electrodes is required and they are sewn to the clothes Two of the EMG electrodes are placed on the muscle bundle 302 Fig 1: EMG acquisition and analysis system diagram 2021 8th NAFOSTED Conference on Information and Computer Science (NICS) fabric (> 99% polyester) were purchased from Yarn dot com Nylon filaments were selected to provide a chemically stable and smooth substrate surface while being highly flexible and stretchable On the other hand, the polyester fabric provided excellent flexibility along with good heat conductivity and electrical insulation Initially, the rGO-coated textile is prepared by dip-coating woven cotton textile into GO (2mg/ml) solution The GO solution would be dried out, and the GO will be thermally reduced at 160°C for hours B EMG electrodes and acquisition circuit Fig 2: Structure of EMG electrodes consisting of graphene sensor, male sensor, and Silver-epoxy (top) and female button to interface with analog acquisiton circuit (bottom) to take differential EMG signal and the third one is used as a reference electrode An EMG electrode is composed of one graphene-based-textile sensor and a metal male button, which are attached to each other by silver epoxy to reduce contact resistance The electrode structure is shown in Figure (top) To connect EMG electrodes to the analog acquisition circuit, female connectors are employed at their interface Figure (bottom) that increases the user’s convenience with quick and easy mating The differential EMG signals are amplified, filtered, and then digitized before being transmitted to the smartphone via Bluetooth connection The EMG signal is then processed, stored, and displayed on the smartphone in real-time The details of these functional circuits are described in the following subsections Fig 4: Block diagram of the proposed EMG acquisition electronic circuit Fig 5: Schematic of the analog acquisition circuit A rGO based graphene sensor-manufacturing Fig 3: SEM images of (a-e) rGO-PE and (f) the illustration how thermal treatment changed the structure of rGO-PE The graphene-based sensor is created by dip-coating nylon and polyester garmnent using the thermally reduced-graphene oxide (rGO) technique The commercial high-purity aqueous solution of Single Layer Graphene Oxide (SLGO) (300800nm, mg/ml), supplied by CheapTubes Inc, The nylon filaments (> 99% polyamide, diameter 500 µm) and polyester The amplitude of EMG signals is small, in the range of tens of micro-volts to some milli-volts, and its frequency range varies between 0-500Hz but the useful EMG signal centralizes around 50 to 150 Hz [10] The EMG electrodes in muscle bundle should be positioned with care to reduce all the artifact noises as well as other noises from environment Two of the EMG electrodes, named MID and END, are placed in the middle and at the end of the desired muscle bundle, respectively The third electrode, called REF, plays the role of a reference electrode and usually placed at the bony or nonadjacent muscular part which is not closed to the target muscle This reference electrode is applied to cancel the inherent motion artifact noises from body signal and the 50Hz power line noise The block diagram of the proposed acquisition circuit is shown in Figure Two EMG electrodes, MID and END, provides differential input signals to the analog acquisition circuit To ensure that the low EMG input signal can be drawn to the acquisition circuit, two buffers are employed according 303 2021 8th NAFOSTED Conference on Information and Computer Science (NICS) to two input paths to introduce high input impedance The differential input signal is then amplified by an instrument amplifier of low-noise and high common-mode-rejection ratio (CMRR) to remove all the mismatches introduced from two EMG electrodes [11], [12] As mentioned above, the useful EMG signal locates around a frequency range of 50 to 150Hz, thus low-pass and high-pass filters are necessary The schematic of analog acquisition circuit is shown in Figure The main duty of high pass filter (HPF) is to cancel the DC offset caused by potential mismatches from instrument amplifier and by the low pass filter Besides, the low frequency artifact noises introduced by the EMG electrodes displacement during measurement process can be mitigated by HPF as well In this paper, a servo-feedback based active HPF is employed to avoid using large capacitor connected in series in main signal path as in a conventional passive HPF The servo feedback is an inverting integrator that is connected in feedback loop between the output of instrument amplifier and its reference input pin, its cutoff frequency is set at 50Hz The HPF is followed by a 1st order low pass filter (LPF) with cutoff frequency of 500Hz that worked as an anti-aliasing filter To minimize power consumption in wireless data transmission, Bluetooth Low Energy module, i.e BLE module NRF 52832 from Nordic Semiconductor, is adopted in this acquisition system The analog EMG signal, after being amplified and filtered by analog acquisition circuit, is then converted into digital signal by a 12-bit analog-digital converter (ADC) preintegrated in BLE module The EMG signal is then transferred to an Android smartphone C Software The digitized EMG data from the acquisition circuit is transferred to the smartphone via BLE connection An Android application (App) is developed to process the received EMG data This application allows users to visualize the EMG signal in real-time on the phone screen as presented in Figure The data is saved in smartphone memory for further analysis such as signal spectrum analysis and muscle fatigue analysis Figure shows an example of Fast Fourier transformation of the EMG signal displayed on smartphone A digital notch filter is also applied in this App to filter out the power line noise instead of applying a complex-and-cost analog filter [13] Fig 6: Real time EMG signal shown in the Android application Fig 7: Fast Fourier transform analysis for EMG signals III E XPERIMENTAL SETUP To evaluate the proposed acquisition system, the EMG signal from the proposed graphene-based-textile sensor is compared to the commercial sensor, i.e Ag/AgCl sensor, through the experiments on different muscles The experiment on biceps muscles are tested as a representation for weak muscle bundles in human body Then, the calf muscles experiment is performed as a representation of stronger muscle groups In the latter experiment, two calf muscles on both legs are tested simultaneously to confirm the possibility of multiple-channel measurement of our proposed system A Measurement setup protocol To validate the quality of the graphene-based-textile sensor, the experiments are realized on biceps muscle on differents subjects and on identical subject for comparison of our proposed sensor and the commercial one, i.e Ag/AgCl To minimize the environmental impact on measured results, a protocol is set and strictly follow by the tester A picture of the experiment protocol is shown in Figure Figure 8a shows the EMG acquisition designed circuit which includes a circular main board connect to an SD card board by a flexible cable Three EMG electrodes are sewed into an handmade armband This armband is designed to wrap tightly in the subject’s arm A thin and elastic fabric layer is used to cover outside of the set-up to keep the whole system stable when tester performs physical activities Fig 8: Monitoring EMG signals from biceps muscle a) Proposed EMG electronic circuit b) Arm flexion c) Arm relaxation For the experiments with biceps muscle, the tester is requested to sit comfortably and steadily, then put his/her arm into the air and flex the arm like Figure 8b every five seconds with a 4-seconds relaxation like Figure 8c between each flexion EMG signal is recorded for graphene sensor base on two different types of fabric (Nylon and Polyester) and Ag/AgCl using proposed electronic circuit For the experiments with calf muscle, the tester is asked to toe stands for second then back to normal standing for 304 2021 8th NAFOSTED Conference on Information and Computer Science (NICS) Fig 9: Measuring EMG signal from calf muscle a) Circuit and knee band a) EMG measurement circuit and placement of electrode b) The system is covered by a thin fabric layer c) Setup in two legs seconds alternating up to five times to ensure a full relaxation of muscle During the test, two separate electronic circuits are used to pick up the signals from two legs A knee band is designed to wrap each leg to keep our sensor in constant contact with the skin Similar to biceps muscle measurement, a thin fabric layer is also used to cover the electrodes and acquisition circuit to make sure the system is stable during experiment The setup is shown in Figure The EMG data from two devices were sent simultaneously to the Android app and saved as two data files respectively IV R ESULTS AND DISCUSSION A Experiment with biceps muscle Figure 10 shows the EMG signal measured from biceps muscle by Ag/AgCl and different types of graphene fabric, i.e Nylon and Polyester fabrics The recording is shown in time domain (Figure 10a) Each active contraction burst of EMG signal corresponds to the moment when the tester flexes his/her arm Otherwise, when the tester in rest stay, there is no signal and only an EMG baseline is observed It can be seen that, the amplitude of the EMG signal acquired by graphene-based-Nylon sensor is higher than that of graphenebased-Polyester as well as the commercial Ag/AgCl In addition, the Fast Fourier Transform is applied to compare the magnitude-frequency domain in the Figure 10b Effectively, the major power spectrum localized in the frequency range of 50 to 150Hz Within this range, The signal obtained from the graphene-based-textile sensor shows higher EMG amplitude than that of Ag/AgCl-based sensor It confirms that the graphense-based-textile sensor offer better performance Indeed, graphene sensor offers a high SNR of 24dB comparing to only 17dB achieved by Ag/AgCl as shown in Table I The SNR is calculated as µsignal (dB), SN R = 20 log µnoise where µsignal and µnoise are the average of RMS values of the received signal during contraction and rest, respectively As mentioned above, a digital notch filter, i.e orders Butterworth IIR, is applied to eliminate the power line interference Fig 10: EMG signals from biceps muscle obtained by Ag/AgCl and graphene-based-textile sensor a) EMG signals in time domain b) in frequency domain TABLE I: EMG SIGNAL PARAMETER COMPARISON Sensor AgAgCl Graphene on Nylon Graphene on Polyester Average amplitude (mV) 6.57 11.67 10.41 SNR (dB) 17 24 24.4 Cross correlation 0.82 0.81 noise, the component of 50Hz is almost cancelled as can be seen in the power spectrum figures (Figure 7, Figure 10b) B Experiment with calf muscle The EMG signals obtained from two legs by graphene sensors are shown in Figure 11 Although the EMG signal from both legs show similar shape, the signal amplitude of right-leg muscle is slightly higher than that of the left leg An EMG pulse corresponds to electric potential generated by muscle cells so it can not be precisely reproduced in 305 2021 8th NAFOSTED Conference on Information and Computer Science (NICS) ACKNOWLEDGMENT This research was supported by National Science Foundation, award number 1853105 and Vingroup Innovation Foundation (VINIF), under the project code DA123 05192019 / year 2019 R EFERENCES Fig 11: Raw EMG signals obtained from graphene sensor from left leg and right leg its previous shape So, in order to quantify the comparison between EMG signal obtained by graphene textile electrodes from left leg and right leg, RMS operation is performed to smooth the signal and still reflect the mean power [8] Figure 12 exhibits the RMS of EMG signal from left leg and right leg It’s obvious that the EMG signal collected from the right leg slightly appeared before the signal from the left leg Besides, the signal amplitude on the right leg is stronger than that on the left leg in nearly all the pulses except for the second one when the pulse from the left leg is 1.8 mV while the right leg is 1.4 mV This could be explained by the fact that the person is right-footed Fig 12: Smoothed EMG signal for graphene sensor from left leg and right leg [1] K Sharmila, T Sarath, and K Ramachandran, “Emg controlled low cost prosthetic arm,” in 2016 IEEE Distributed Computing, VLSI, Electrical Circuits and Robotics (DISCOVER) IEEE, 2016, pp 169–172 [2] O M Giggins, U M Persson, and B Caulfield, “Biofeedback in rehabilitation,” Journal of neuroengineering and rehabilitation, vol 10, no 1, p 60, 2013 [3] J Heaffey, E Koutsos, and P Georgiou, “Live demonstration: Wearable device for remote emg and muscle fatigue monitoring,” in 2015 IEEE Biomedical Circuits and Systems Conference (BioCAS) IEEE, 2015, pp 1–5 [4] M A Lopez-Gordo, D Sanchez-Morillo, and F P Valle, “Dry eeg electrodes,” Sensors, vol 14, no 7, pp 12 847–12 870, 2014 [5] P Laferriere, E D Lemaire, and A D Chan, “Surface electromyographic signals using dry electrodes,” IEEE Transactions on Instrumentation and Measurement, vol 60, no 10, pp 3259–3268, 2011 [6] D Brunelli, A M Tadesse, B Vodermayer, M Nowak, and C Castellini, “Low-cost wearable multichannel surface emg acquisition for prosthetic hand control,” in 2015 6th International Workshop on Advances in Sensors and Interfaces (IWASI), 2015, pp 94–99 [7] F Awan, Y He, L Le, L.-L Tran, H.-D Han, and L P Nguyen, “Electromyography acquisition system using graphene-based e-textiles,” in 2019 International Symposium on Electrical and Electronics Engineering (ISEE), 2019, pp 59–62 [8] P Konrad, “The abc of emg,” A practical introduction to kinesiological electromyography, vol 1, no 2005, pp 30–5, 2005 [9] A Gonzalez-Mendoza, A I Perez-SanPablo, R Lapez-Gutierrez, and I Quinones-Uriostegui, “Validation of an emg sensor for internet of things and robotics,” in 2018 15th International Conference on Electrical Engineering, Computing Science and Automatic Control (CCE), 2018, pp 1–5 [10] A Goen and D C Tiwari, “Review of surface electromyogram signals: Its analysis and applications,” International Journal of Electrical and Computer Engineering, vol 7, no 11, pp 1429 – 1437, 2013 [11] J Gomez-Clapers and R Casanella, “A fast and easy-to-use ecg acquisition and heart rate monitoring system using a wireless steering wheel,” IEEE Sensors Journal, vol 12, no 3, pp 610–616, 2012 [12] A Shafti, R B Ribas Manero, A M Borg, K Althoefer, and M J Howard, “Embroidered electromyography: A systematic design guide,” IEEE Transactions on Neural Systems and Rehabilitation Engineering, vol 25, no 9, pp 1472–1480, 2017 [13] D T Mewett, K J Reynolds, and H Nazeran, “Reducing power line interference in digitised electromyogram recordings by spectrum interpolation,” Medical and biological engineering and computing, vol 42(4), no 4, pp 524–531, 2004 V C ONCLUSION In this paper, a novel easy-to-use system intended for the flexible and wearable monitoring of EMG signal by using graphene-based-textile sensor has been presented The obtained signals from biceps and calf muscles are evaluated A better performance is observed for graphene-based-textile sensor in comparing with the commercial sensor The experiments analysis shows that the reusable rGO-coated fabric is a promising alternative to commercial sensor in muscular activity monitoring applications 306 ... the graphene- based- textile sensor shows higher EMG amplitude than that of Ag/AgCl -based sensor It confirms that the graphense -based- textile sensor offer better performance Indeed, graphene sensor. .. SETUP To evaluate the proposed acquisition system, the EMG signal from the proposed graphene- based- textile sensor is compared to the commercial sensor, i.e Ag/AgCl sensor, through the experiments... treatment changed the structure of rGO-PE The graphene- based sensor is created by dip-coating nylon and polyester garmnent using the thermally reduced- graphene oxide (rGO) technique The commercial high-purity