Robust Control of Hand Prostheses from Surface EMG Signal for Transradial Amputees

Abstract

This paper investigates the difficulty in gaining robust control of hand prostheses by the Surface Electromyogram (sEMG) of transradial amputees under dynamic force levels because these changes can create a significant effect on robust controlling of the prostheses. A set of attributes has also been proposed to lessen the effect of force level changes on the prosthetic hand users which is controlled by amputees. To accomplish this task, at first the signal is pre-processed to abolish noise and artefacts from the raw-sEMG signals and then extracts features. Features are extracted from three specific domains: time, spectral and wavelet domain. For farther analysis, wavelet packet and entropy-based features have also been extracted. Finally, for classification purpose state-of-the-art classifiers such as Support Vector Machine (SVM), Decision Tree (DT), Linear Discriminant Analysis (LDA) and Quadratic Linear Discriminant Analysis (QLDA) have been used and compared. To optimize the hyper-parameters of classifiers, Bayesian optimization algorithm has been used. Our recommended system is verified through openly accessible EMG database and results relate with the proposed system. Classification was done under four well known classifier which are DT, LDA, QLDA and SVM respectively and their accuracy is calculated for both feature and signal level.