NON NEGATIVE MATRIX FACTORISATION FOR IDENTIFICATION OF EMG FINGER MOVEMENTS

Abstract

Electromyography (EMG) signals are becoming important in many applications, including clinical/biomedical, prosthesis or rehabilitation devices, human machine interactions, etc. This work present classification of Surface Electromyography (sEMG) using four different methods, namely, Artificial Neural Network (ANN), Discriminant analysis, Multi-Support Vector Machine (m-SVM) and K-Nearest Neighbour (KNN) method and compares the accuracy of classification of these methods. Also, all the four methods use two methods, namely the nonnegative matrix factorization (NMF) and principal component analysis (PCA) for dimensionality reduction of data. MATLAB simulations show that ANN classifier gives the best accuracy among the four classifier used in this work. The percentage accuracy of ANN classifier is 95% using NMF method and 84.5% using PCA method.