Surface Electromyography (EMG) Signal Processing, Classification, and Practical Considerations

Abstract

Electromyography (EMG) is the process of measuring the electrical activity produced by muscles throughout the body using electrodes on the surface of the skin or inserted in the muscle. EMG pattern recognition based myoelectric control systems typically contain data pre-processing, data segmentation, feature extraction, dimensionality reduction, and classification. The real challenge for prostheses and gesture recognition interfaces are the dynamic factors that invoke changes in EMG signal characteristics. As a consequence of these factors, model inaccuracies are produced between the training phase and practical use. In this chapter, state-of-the-art EMG signal processing and classification techniques that address these dynamic factors and practical considerations are presented. Hands-busy conditions and cross-user classification models that present additional challenges for gesture recognition tasks are also explored. Finally, directions for future research are outlined and discussed.