Design and Development of a Wireless Surface EMG Band for Real-Time Muscle Signal Monitoring

Abstract

Surface electromyography (sEMG) is a non-invasive method to monitor muscle activity. Due to its versatility and suitability for wearable systems, this technology can be applied in a wide range of fields, including rehabilitation, prosthetics, and humanmachine interaction. This bachelor’s thesis aims to design and develop a wireless sEMG armband for real-time monitoring of muscle activity, including movement classification of three key movements. The main aspects are hardware construction, wireless transmission, signal processing, and a graphical user interface (GUI). The development of each part involved an iterative work process. A multi-channel bipolar method was used for the sEMG recordings. The electrodes were made of stainless steel with dimensions 15×8×0.5 mm. Three measurement channels were used, with the signal from each amplified and filtered using an instrumentation amplifier, an RC band-pass filter, and an operational amplifier. The signals are then sampled and wirelessly transmitted via ESP-NOW to a computer. Each sample is processed using custom root mean square (RMS) and continuous wavelet transform (CWT) functions and displayed in a GUI. Using the extracted features, three movements are classified using linear discriminant analysis (LDA): rest, wrist extension, and flexion. Each part of the system was tested and verified individually. The electrodes exhibited characteristics similar to those of commercially available Ag/AgCl wet electrodes. However, the amplification and cutoff frequencies differed from the calculated and desired values. The lower cutoff frequency varied between 13 and 15 Hz, and the upper between 299 and 350 Hz across the three channels. The latency for the wireless transmission is on average 20.8 ms, with 0% packet loss. Both the RMS and CWT functions were deemed effective and consistent with the theoretical expectations. The movement classification resulted in an accuracy of 93.9 %. Despite isolated tests indicating that each part works, the system as a whole lacks adaptability and robustness. For instance, the movement classification heavily relies on manual calibration. To make the band viable for practical use, several improvements are required. However, it is concluded that the band achieves its objective of recording and monitoring muscle activity in real time and classifying key movements, while serving as a proof-of-concept for a wireless sEMG armband.