Integrating Wearable Textiles Sensors and IoT for Continuous sEMG Monitoring

Abstract

Surface electromyography is a technique used to measure the electrical activity of muscles. sEMG can be used to assess muscle function in various settings, including clinical, academic/industrial research, and sports medicine. The aim of this study is to develop a wearable textile sensor for continuous sEMG monitoring. Here, we have developed an integrated biomedical monitoring system that records sEMG signals through a textile electrode embroidered within a smart sleeve bandage for telemetric assessment of muscle activities and fatigue. We have taken an "Internet of Things"-based approach to acquire the sEMG, using a Myoware sensor and transmit the signal wirelessly through a WiFi-enabled microcontroller unit (NodeMCU; ESP8266). Using a wireless router as an access point, the data transmitted from ESP8266 was received and routed to the webserver-cum-database (Xampp local server) installed on a mobile phone or PC for processing and visualization. The textile electrode integrated with IoT enabled us to measure sEMG, whose quality is similar to that of conventional methods. To verify the performance of our developed prototype, we compared the sEMG signal recorded from the biceps, triceps, and tibialis muscles, using both the smart textile electrode and the gelled electrode. The root mean square and average rectified values of the sEMG measured using our prototype for the three muscle types were within the range of 1.001 ± 0.091 mV to 1.025 ± 0.060 mV and 0.291 ± 0.00 mV to 0.65 ± 0.09 mV, respectively. Further, we also performed the principal component analysis for a total of 18 features (15 time domain and 3 frequency domain) for the same muscle position signals. On the basis on the hierarchical clustering analysis of the PCA's score, as well as the one-way MANOVA of the 18 features, we conclude that the differences observed in the data for the different muscle types as well as the electrode types are statistically insignificant.

Keywords: IoT-integrated textile sensor; electrode position; sEMG; smart wearable; textile sensor.

Figure 4

The embroidery process of the textile electrode development, on the textile bandage sleeve of the sEMG monitoring system.

Figure 5

3D-printed final proof-of-concept prototype of the sEMG device, with its housing.

Figure 1

Block diagram of the monitoring device.

Figure 2

(a) Myoware muscle sensor unit, (b) node MCU microcontroller unit.

Figure 3

Schematic diagram of the monitoring system.

Figure 6

(a) sEMG signals recorded from biceps, using gel(functional) electrode, (b) conductive hybrid thread-based embroidered electrodes [30], (c) the duplicate measurement.

Figure 7

(a) sEMG signals recorded from triceps using gelled electrode, (b) hybrid thread-based embroidered electrodes (c) the duplicate measurement.

Figure 8

(a) sEMG signals recorded from tibialis using gelled (functional) electrode, (b) hybrid thread-based embroidered electrodes, (c) the duplicate measurement.

Figure 9

3D plot of the normalized input data used for PCA. The normalized intensities of the pareto-scaled features obtained from the sEMG data of different muscle positions, namely tricep, bicep and tibialis muscles, are shown in the green, red and blue traces, respectively.

Figure 10

Biplot plot of the sEMG data for textile and gelled (functional) electrodes, compared for tricep (●), bicep (●), and tibialis (●) muscles. The score of the data measured by gelled (functional) and textrode (and its duplicate), which are shown as solid circles; component features corresponding to the first two loading vectors are shown as a line ending with a dot.

Figure 11

The hierarchical clustering is based on the PC1 & PC2 scores of the nine data: X1, X2, X3 (Triceps), X4, X5, X6 (Biceps), X7, X8, X9 (Tibialis). The analysis has identified two significant clusters bounded by two rectangles, at the significance level of 95% or α = 0.05. We observe that X1–3, which belongs to triceps is, as expected, grouped together (right cluster). But we also observe that the X2 and X9, which belonged to bicep and tibialis data, respectively, are also grouped with triceps. This implies that the variance observed among data (score values) is not very significant and the clustering algorithm could not differentiate each muscle type as separate clusters. Both the AU and BP based p-values for each cluster is shown in red and green fonts.