Motor State Classification based on Electromyography (EMG) Signals using Wavelet Entropy and Neural Networks

Neuromuscular system works based on communication between nerves and muscles. Capturing neuromuscular activities is employed in many application domains, ranging from remote-robotic control, rehabilitation, remote surgery, etc. The ability to accurately decode the intention determines the reliability of the application. This paper proposes a motor state classification technique based on Electromyography (EMG) signals. Current research in the general domain of EMG signal classification has generated good results. However, the existing techniques require high computing resources or are non-real-time in nature, making them impractical for real-world application. In this paper, the authors propose the use of wavelet entropy and neural networks for classification of four major hand movements, forward, reverse, raise and lower. The results ranging from 91.9% to 93.5% show promise in the extension of the proposal for further classification