Decoding Muscular Signals: Machine Learning Approaches to EMG Classification

Electromyography (EMG) classification using machine learning techniques has gained significant attention in recent years due to its applications in various aspects such as prosthetic control, gesture recognition and muscle health monitoring. In this study, we explore the applications of both Machine Learning and Deep Learning techniques for the classification of EMG signals. In this work, a new CNN and DFNN models is proposed for this purpose achieving high accuracy with low computation time. Additionally, several machine learning algorithms are evaluated for EMG classification, including Random Forest, KNN, AdaBoost, Decision Tree with and without cross-validation were employed. Moreover, we investigate the impact of class balance on the performance of these models. Model selection and hyperparameter tuning are conducted to optimize the performance. The models are assessed based on accuracy, precision, recall and [Formula: see text]-score. The best results are obtained using random forest with an accuracy of 99.81% while the proposed CNN model achieved an accuracy of 99.61%. Experimental results proved the efficiency of the proposed work compared to other state-of-the-art works.