Investigation of Optimum Pattern Recognition Methods for Robust Myoelectric Control During Dynamic Limb Movement

Abstract

The control of upper limb prostheses based on surface electromyogram (EMG) pattern recognition has long been the focus of many researchers as an important clinical option for amputees. More recently, it has been shown that changes induced during use, such as changes in limb position and performing dynamic activities, can have a substantial impact on the robustness of EMG pattern recognition. This work investigates whether there are alternative EMG features and classifiers which can outperform the commonly used time domain (TD) features and linear discriminant analysis (LDA) classifier in the context of limb positional changes and performing dynamic activities of daily living. A variety of EMG feature combinations and popular classifiers are compared in this study. The bases of comparison are classification accuracy and class separability. The results showed that adding Willison amplitude (WAMP) feature to the commonly used TD feature set combined with LDA classifier reduces the averaged absolute classification error by 1.4%.