Performance of various EMG features in identifying ARM movements for control of multifunctional prostheses

Abstract

In this study, we evaluated classification performance of electromyography (EMG) four time-domain features and autoregressive model features and their combination in identifying 11 classes of arm and hand movements in both able-bodied subjects and amputees. Our results showed that using three time-domain features could achieve similar classification accuracy as using four features. Using AR model coefficients as EMG features, a six-order AR model might be optimal. For the evaluation of performance of EMG pattern recognition in identifying various movements, the amputees should be used. The outcomes of this study may aid the future development of a practical multifunctional myoelectric prosthesis for arm amputees.