Electromyography (EMG) based Classification of Finger Movements using SVM

Myoelectric control prostheses hand are currently popular developing clinical option that offers amputee person to control their artificial hand by analyzing the contacting muscle residual. Myoelectric control system contains three main phase which are data segmentation, feature extraction and classification. The main factor that affect the performance of myoelectric control system is the choice of feature extraction methods. There are two types of feature extraction technique used to extract the signal which are the Hudgins feature consist of Zero Crossing, Waveform Length (WL), Sign Scope Change (SSC) and Mean Absolute Value (MAV), the single Root Mean Square (RMS). Then, the combination of both is proposed in this study. An analysis of these different techniques result were examine to achieve a favorable classification accuracy (CA). Our outcomes demonstrate that the combination of RMS and Hudgins feature set demonstrate the best average classification accuracy for all ten fingers developments. The classification process implemented in this studies is using Support Vector Machine (SVM) technique.