Machine and Deep Learning-Based Detection of Forearm Muscle Contraction Onset Using EMG Signals

Abstract

Accurate identification of muscle contraction events plays a crucial role in the development of emerging technologies in the field of bioelectrics. In this study, a hybrid approach based on surface Electromyography (sEMG) signals and deep learning algorithms is proposed for the detection of the onset and offset of forearm muscle contractions. Muscle activity data were collected from six healthy volunteers using silver/silver chloride (Ag/AgCl) electrodes and Arduino-based signal amplification modules. Following preprocessing steps, including Butterworth filtering and noise removal, the signals were segmented into 5-second windows. Subsequently, ten time- and frequency-domain features were extracted from each window. In the next stage, data classification was performed using three algorithms: 1) a Multilayer Perceptron (MLP), 2) a Gaussian Mixture Model (GMM), and 3) a Convolutional Neural Network (CNN). The results demonstrated that the CNN classifier achieved superior performance, with an accuracy of 94.21%, compared to GMM (76.15%) and MLP (68.81%). These findings highlight the high effectiveness of deep learning–based methods in the accurate detection of muscle contraction events. The limitations of this study include the small sample size and the lack of full control over the participants’ physiological conditions. Therefore, future research is recommended to employ larger datasets and more realistic movement conditions.