Data acquisition and feature extraction for classification of prehensile semg signals for control of a multifunctional prosthetic hand

Official URL: http://192.168.1.20/record=b1225683 (Table of Contents)

This study focuses on the SEMG (surface electromyography) signals that carry the valuable information of the neuromuscular activity of a muscle and are utilized in the man-machine interfaces such as multi-functional prostheses. The SEMG signals measured from four different muscle groups of the forearm are weak, sophisticated and very sensitive to ambient noise. The first stage of this study is hardware design and implementation for the SEMG measurement. The fundamentals of the design are mainly based on the specifications of the SEMG signal and the factors that affect the signal quality. The second purpose of the thesis is applying various methodologies to the recorded SEMG signal to give meaning to its nature to be used in the further processes. The raw EMG signals have nonlinear characteristics and present useful information if they are quantified. For this purpose, various signal processing methods are applied to the SEMG signal to acquire useful information, features. Features of the signal are extracted to be used for classification of prehensile motions of multi-functional prosthetics. In this part, many algorithms that have been employe as feature extraction methods are compared with respect to their classification performance.