Wavelet based neuro-fuzzy classification for EMG control

High accuracy of multiple degrees of freedom in prosthetic control is hard to obtain because uncertainty exists between different movements. Neuro-fuzzy technology is suitable to deal with such problems. We adopt a wavelet based neuro-fuzzy approach to classify EMG signals for movement recognition in order to decrease classification error. EMG signals are analyzed by wavelet transform, and feature vectors are constructed by SVD transform from wavelet coefficients for further movement recognition. A neuro-fuzzy network is designed as classifier, and its initialization and training are also involved. Comparison results for this method and traditional ones are provided to show its efficiency. High recognition and reliability are achieved in preliminary experiments.