The Application Research Of Principle Component Analysis And Nonlinear Method In Surface Electromyography

The surface electromyography (sEMG) signal, which is noted by the surface electrodes, is the electrical manifestation of the neuromuscular activation associated with contracting muscles. It varies with different action, and based on this, the dissertation emphasizes on how to extract the feature from EMG, in order to achieve higher accuracy of action pattern recognition.Above all, the physiological theory of sEMG signal and its applications are introduced. At the same time, the existing methods of identifying sEMG are discussed. This knowledge is the foundation of next study.Afterwards, three new methods are innovated to study pattern recognition of different actions. The first one is the method that combines feature extraction based on multi-AR model and PCA (principle component analysis) with pattern classifier based on support vector machine, results in 98.89 % correct rate for pattern classification of six movements ; The second one is the method that combines feature extraction based on MSPCA (multi-scale PCA) with bayes classifier,results in 99.44 % correct rate; The third one is the method that combines feature extraction based on wavelet coefficient and kernel function with bayes classifier,results in 100 % correct rate. Therefore, with the advancement of method for feature extracting, the correct rate of pattern classification is higher.Finally, nonlinear dynamic method is employed in this paper to do the research of sEMG's nonlinear character. Then the theory of phase space reconstruction on nonlinear time series is expatiated,with which the time delay and the optimum-embedding dimensions are obtained to reconstruct the phase space of surface EMG, further more, the largest Lyapunov exponent, correlation dimensions, and approximate entropy are calculated. The results confirm that the sEMG signal has nonlinear deterministic component and it is correlated with chaos, leading to our further recognition and understanding of EMG's nonlinear character.PCA and nonlinear methods are applied in sEMG study through all the paper, and each of them results in good performance, which proposed new methods and different thinking for improving the research of ASEMG.