Hybrid approach for myoelectric control of multi-finger movement classification

Abstract

Identification of correct multi-finger movement class remains a difficulty in a myoelectric prosthetic hand. This is because only a small amplitude of electromyography (EMG) signal was produced by this multi-finger movement. Hence, powerful classification is needed to solve this problem. Support Vector Machine (SVM) is a classification method that has been widely used in classifying multi-finger movement. However, SVM only able to generate solution of multi-finger classification based on non-optimal default parameter. Hence, the objective of this research is to propose hybridization of Grey Wolf optimizer (GWO) with SVM namely hybrid GWO-SVM approach to enhance multi-finger movement classification. The multifinger movement dataset used in this study was from Khusaba et al. (2012) downloaded from free public database in raw forms. The data were generated from two surface EMG channels patched on the remaining limb using Delsys DE 2.x series EMG sensors. The generated EMG signal was then amplified using Delsys Bagnoli8amplifier and sampling using A 12-bit analogue-to-digital converter (National Instruments, BNC- 2090) at 4000Hz. Both amplified and sampling processes were completed using Delsys EMGWorks Acquisition software. Next, pre-processing and feature extraction are important for the achievement in EMG analysis and control and by utilizing the feature extraction process, we can reduce the computational cost of a multifunction myoelectric control system. Furthermore, Hudgins feature set and Root mean square (RMS) feature extraction method were also employed to produce optimal features. The results showed that the proposed hybrid GWO-SVM approach has improved the classification accuracy, sensitivity, and specificity by 1.52 %, 14.22 % and 18.77 % respectively. Hence, the proposed hybrid approach can help in improving the performance of prosthesis hand for prosthetics people.