Wrist Motion Recognition by Using Electromyographic Signals

Abstract

Wrist motion classification is a very common research topic in scientific study. However, wrist motion recognition of the surgeon is often neglected in the robot-assisted surgery or surgical training. Therefore, the objective is to develop a classification method to recognize wrist motion of the surgeon. In order to do that, we present a linear discriminant analysis (LDA) algorithm involving surface electromyography (sEMG) signals to evaluate the motions in this paper. Firstly, sEMG signals are collected by using a MYO armband which can be worn on the forearm of a subject. Root-mean-square (RMS) and waveform length (WL) feature are extracted from the sEMG signals and then those features are regarded as input of the LDA to train the classifier. As a result, we can obtain a classifier to recognize four kinds of wrist motions. Classification experiment is performed by two subjects. The experimental results have been demonstrated by using the proposed approach and it is shown that the accuracy of wrist motion by using RMS feature is higher than that of by using WL feature.