Research Of Pattern Recognition Of SEMG Based On High Order Statistics

Surface electromyogram (sEMG) is a complex electric signal which is depend on the contracting muscles beneath the skin. It can be recorded non-invasively from the skin by surface electrodes. SEMG which are from different action patterns of upper limbs are different. Action patterns can be recognized by analyze sEMG of upper limbs. And this method can be used in the control of human-robotic arms.This paper use high order statistics in the classification of surface electromyogram (sEMG). The approach utilizes bispectrum analysis on sEMG signal classification to classify six primitive motions,this six pattern recognition including varus, ectropion, hand grasps, hand extension, upwards flexion and downwards flexion. In the Pattern Recognition of sEMG before, the sEMG signal is assumed to be Gaussian, linear and stationary. But the truth of sEMG is often not satisfy the assumption above. According the research on sEMG before we know that, the level of non-Gaussianity of sEMG signal recorded in muscular forces below 25% of maximum voluntary contraction (MVC) is significant. So in order to obtain more information of sEMG and better recognition accuracy of pattern recognition of sEMG,we use bispectrum analysis combine with principal component analysis and bispectrum slice to make classification research on EMG.We get good recognition results from this two methods.