Exploring Machine Learning Algorithms for Optimized Muscle Activity Detection Using sEMG Signals: A Systematic Review

sEMG signals are considered important in development of prosthetics, clinical diagnosis and rehabilitation sciences. The most important feature in signal acquisition is muscular-activity timing that determines muscle onset/offset timing representing indicating muscle activity. Numerous research propositions are devised, although the standard method is not completed yet. The aim of this systematic study is to review the reliability of multiple machine-learning-based approach for detection of muscle activity through sEMG signals to compare the performance with earlier study. IEEE Xplore, Google Scholar and Science Direct websites were preferred to retrieve resources for foundation of our research publications between 2010 and 2023 by implementation of suitable keywords as ‘‘muscle activity and Machine Learning .’’ With proper protocol of careful screening, 23 publications were chosen in our selected criteria in this written systematic review. The work scrutiny reveals that provided Machine Learning (ML) techniques comprised of Artificial Neural Network (ANN), Support Vector Machine (SVM), K-nearest Neighbor, and TKEO. Artificial Neural Networks were reflected to demonstrate the major accuracy of 95% proceeding by SVM providing 92% and TKEO generating 85%. These techniques are majorly used for proposing protocol examinations involved in EMG parameters. Values of parameters such as accuracy, (SNR), threshold, error were uprooted from the studies and relevant conclusion were also made through consideration of comparisons provided in data statistics. Eventually, with this provided systematic review, a brief compilation of the studies was executed that comprises of how Machine Learning (ML) methods have been implemented for the efficient detection of muscle activity.