Surface Electromyographic Hand Gesture Signal Classification Using a Set of Time-domain Features

Abstract

The surface electromyographic (sEMG) signal-based hand gesture recognition system has been widely adopted for the development of prosthetic control, robotics, and surgical systems. However, it is a challenging task to extract distinguishable features from the sEMG signal for accurate recognition of the gesture class. In this work, a set of timedomain features (SoTF) are extracted from each channel of the sEMG signal for effective recognition of the gesture class. The proposed SoTF is a combination of average, standard deviation, and waveform length features extracted from each channel. The classification accuracy using the SoTF is compared for three different classifiers such as k-nearest neighbors (KNN), support vector machine (SVM), and random forest (RF) on 52 gesture classes of NinaPro DB1 dataset. Variations in parameters of the classifiers are also analyzed to obtain the best classifier. Experimental results show that the SoTF with RF classifier achieves superior performance compared to the state-of-the-art techniques