The Application Of Time-frequency And Nonlinear Time Series Analysis In EMG Signal's Processing

Through combining the physiological characteristic of muscle contraction, this paper uses the time-frequency analysis and nonlinear time series analysis to process the athletes' surface EMG signals, as well as to evaluate effectively the athletes' training level and athletic state.This paper firstly discusses the mechanism of EMG and its attribution in Sport Biomechanics. The current research status of sEMG is introduced as well. Then the paper explains the meaning of this research and the experiment process that acquires sEMG signals from sprinters and butterfliers.Because surface EMG is non-stationary, time-frequency analysis based on reassignment method is introduced, which proves to have little interference and good resolution both in time and frequency domain. The theory of Empirical Mode Decomposition (EMD) as well as Hilbert spectrum is also presented. Then by combining reassignment method and EMD, the surface EMG signal from sprinters is analyzed. The experiment results show that Intrinsic Mode Function (IMF) can pick out the sEMG's dominant 50-150Hz frequency range, and explain the way of motor units' recruitment of slow and fast conducting muscle fibers. To get a quantified parameter to evaluate sprinters' training level and athletic state, energy ratio between different IMFs is presented, which is also explained from the view of muscle's physiology theory.EMG signal is nonlinear from its physiological viewpoint. This paper introduces surrogate data method and uses it to prove that athlete's EMG signal is nonlinear; in addition, it proves that biceps brachii is weakly nonlinear, but rectus femoris and vastus medialis are strongly nonlinear, which both are explained with muscle's physiological theory. Then the theory of phase space reconstruction on nonlinear timeseries is studied, with which the time delay and the optimum-embedding dimension are obtained to reconstruct the phase space of surface EMG, based on which the largest Lyapunov exponent is calculated with an improved Lyapunov exponent algorithm for small data sets, whose results confirm the surface EMG signal is correlated with chaos. Then the correlation dimension of surface EMG is acquired, which confirms the conclusion of the above two sections. At last, a method of fractal dimension based on spectrum analysis is used to evaluate sprinters' training level and athletic state.