Comparative Analysis of Hand Gesture Classifiers Using EMG Signal and STFT-CNN

Abstract

Electromyogram (EMG) signals are used to detect different hand gestures of a person. In this study, a Convolutional Neural Network (CNN) based feature extraction-based method is used, by which features of processed signals are extracted using a CNN model, and the features are then classified using a Machine Learning classifier. Three-hand gestures models were developed during the study, ten, seven, and four hand gestures are used to develop three models and channel wise analysis is also performed for finding of dominant channel location. These models may be helpful in the development of a prosthetic hand for patients with upper limb disabilities or a robotic hand for industrial purposes.