An Efficient Approach for Feature Selection of SEMG Signal

Abstract

This paper introduces an approach to obtain the feature vectors of surface electromyography (sEMG) signal based on Hilbert Huang transform (HHT). An adaptive segmentation method that could effectively select appropriate intrinsic mode function (IMF) is proposed. With the features gathered by using the energy of one channel signal, we also provide an optimized strategy based on experiments and experiences to increase the recognition rate of hand-motion patterns. The results from SVM neural networks classifier are presented to support this approach.