Prediction of joint dynamics from electromyography signals using artificial neural networks

Abstract

Wheelchair related issues are attracting more and more attention as an estimated population of 1,200,000 in the United States rely on wheelchairs for their mobility (Philips and Nicosia, 1990). Currently, there are three major wheelchair research areas: (1) the metabolic and cardiopulmonary characteristics of the wheelchair user such as oxygen consumption and heartrate, (2) the design and construction of wheelchairs, and (3) the kinematic features during the wheelchair propulsion. The studies of kinematic features while propelling a wheelchair have focused on the movement of the joints and body segments, which are the outcomes of muscle activities. The purpose of this study was to conduct a wheelchair propulsion experiment to (1) build an artificial neural network (ANN) model to correlate electromyography signals (EMGs) to kinematic features of wrist, elbow, and shoulder joint using small data sets, and (2) use the trained ANN model to predict joint dynamics from known EMG data.