Feasibility Study of Detecting the Impact of Caffeine, and Diet on Hand Gestures Classification by sEMG Signals

When using a set of data collected under real conditions for machine learning applications, it is very important to consider all the circumstances that may affect the quality of the acquired signals. The influence of fasting, caffeine, and large meal consumption on surface electromyography (sEMG) signals collected from healthy subjects is investigated in this paper. Four Myoware sensors are used to collect sEMG signals from a healthy male subject while performing four groups of hand rehabilitation exercises. In this paper, three classification algorithms specified as k-Nearest Neighbors (kNN), Random Forest, and Support Vector Machine (SVM) are tested and compared. The hand gesture classification comparison results prove that the SVM approach achieves the best classification accuracy equal to 91.66% while using the signals collected when the subject is fasting.