Comparison of SVM and ANN Classifier for Surface EMG Signal in Order to Improve Classification Accuracy

Abstract

Surface EMG is significant signal initiating from human body while doing various movements. This can be exploited for various applications like movement classification, diagnosing neuromuscular disorders, prosthetic control and many more. Several researchers are trying to provide solutions for tackling this problem in the form of improving acquisition circuit for surface EMG signal, increasing the density of sensors during acquisition process, extracting novel features which could give more information and so on. One of the crucial stages while analyzing surface EMG signal is selection of feature sets and classification algorithm. In present work the authors tried different time domain feature sets and their combinations to improve classification accuracy. The ANN and SVM classifiers are compared as these two are very popular among the researchers. It was observed that a combination of feature sets improves classification accuracy but response time is increased. It was also observed that ANN performed better in terms of classification accuracy as compared to SVM. The present study explains the optimized solution for the aforesaid problem.