A Research Of Human Upper Limbs Surface Electromyography Signal Analysis Methods

Surface electromyography is a faint and complex biomedical signal, which comes from contraction of muscle and is controlled by neurons. Hence, the analysis of surface electromyography could reflect the variation controlling tendency of neuromuscular, consequently, it can be applied to the study of rehabilitation robot system.This thesis is based on the 985 project research and application of rehabilitation robot system key technology. The main research involves building surface electromyography signal acquisition system, surface electromyography signal’s pretreatment, de-noising and feature analysis. This thesis build surface electromyography signal acquisition system, solved surface electromyography signal de-noising problems and used the method of fractal theory analysis on the surface electromyography research all successfully.The surface electromyography signal collection system this thesis designed is a physiological signal acquisition device named FlexComp Infiniti with multi-channel and multi-parameter. And Biograph Infiniti software is used to analysis the collected signal. Through large number of the signal acquisition experiments,a conclusion can be got that this system work effectively.During the process of surface electromyography signal acquisition, there are many interference signals mixed with wanted signals,so signal denoising is needed. In this thesis, the AR model method is first carried out on the surface electromyography filtering denoising, and get the desired effect. Secondly, wavelet analysis method is used to signal thresholding denosing. Experimental results have proved that the combination filtering methods based on AR model and based on wavelet transform can well achieve the removal of surface electromyography signal noise.After the de-noising study on surface electromyography signals and considering fractal theory and calculation methods, the fractal dimension is the valid expression parameter of the fractal object. Combined with fractal dimension calculation method and multiple linear regression method, presents a more consistent surface EMG as a calculated fractal dimension. Experimental results show that the proposed method of surface EMG in different actions extracting characteristic dimension can well distinguish different actions.By the research of this paper, we proposed a series of algorithms and implementing approaches of surface electromyography signal acquisition. I will continue to study the various analysis and processing methods to find which one is more suitable for surface electromyography signal in future.I try to use intelligent classifier on different actions for pattern classification simultaneously and apply to rehabilitation robot system finally.