The surface EMG signals which are gathered by experiments are analysed and dealt by means of the method of chaos and fractal theory, and the chaos of the surface EMG signals and the characteristic of fractal in the process of the movement is analyzed by means of reconstructing phase space in this paper. It is researched that, the surface EMG signals have positive Lyapunov exponent, and show certain chaos features. It is found that myoelectric signals of differernt motions have different cluster distributions by calculating the surface EMG signals’fractal dimension, and also can recognise and analyze the distilled characterless value——ftactal dimension by all kinds of classifier, obtaining better identify effect, especially, the identify quotiety to the eight action using the classifier based on Support Vector Machine theory and Bayes Decision theory all can be above 90%.