Classification of the surface EMG signal using RQA based representations

Feature extraction is a key element of pattern recognition for myoelectric control. In this paper, recurrence plots and recurrence quantification analysis (RQA) are used as the feature extractor for surface EMG signals. For eight different hand motions, two-channel EMG signals are recorded. Ten individual RQA parameters are calculated for each channel of EMG signals. With different combinations of individual RQA parameters, a set of feature vectors with dimensions varying from 2 to 20 are obtained. The feature vectors are used as the input to a BP neural network for motion classification. Experimental results show that with appropriate selections of feature vectors, the motion classification algorithm achieves desirable accurate rate.