Comparison of SVM an/kNN classifiers for palm movements using sEMG signals with different features

Abstract

The human-machine interface plays a major role in the development of the prosthetic arm which acts as an immediate rehabilitation for the amputee. Electromyogram (EMG) signals which are signals acquired from muscles require high accuracy in detection, preprocessing, feature extraction and classification which is a challenging task. The main focus of this paper is on how to improve the accuracy by using low cost electrodes so that the overall prosthesis cost can be lowered. Two different classification techniques, Support Vector Machine (SVM) and k- Nearest Neighbor (kNN) are employed and the results are compared to determine which method gives better accuracy. In this paper, classification and analysis is done on surface Electromyogram (sEMG) signals acquired from muscle sensor v3 and datasets available online. The movements to be classified are cylindrical, spherical and lateral. For analysis of the signals , four level wavelet decomposition was used and features such as Standard Deviation (SD),Waveform Length (WL) and Root Mean Square (RMS) were extracted and compared.