Multiple Features Fusion System for Motion Recognition

Abstract

Surface EMG signal is a signal source that can reflect the movement state of human muscles accurately. However, there are still problems such as low recognition rate in practical applications. It is necessary to study how they can be exploited effectively for a more accurate extraction. The paper combines two time domain features and nonlinear feature to get the feature vector for subsequent pattern recognition. The paper chooses the generalized regression neural network (GRNN) as the classifier for hand motion pattern recognition. The proposed method in this paper not only realizes the feature extraction of signals, but also ensures the high classification accuracy. The feature, RMS-SampEn-WL, obtains the highest recognition rate above 97% compared with the two time features. The new sEMG feature is effective and suitable for hand motion pattern recognition. Finally, we hope to establish a robust recognition system based on sEMG.