A Study on Instantaneous Median Frequency of Surface EMG Signals Based on Time-frequency Distributions

Objective Median frequency of EMG signal is a significant physiological parameter of the neural muscular system. Due to the non stationarity of EMG signal, the methods such as periodogram and parameters model may not be suitable for frequency analysis of EMG signal. We use the time frequency analysis method to detect the instantaneous frequency components of EMG signal. Method The concept of instantaneous median frequency surface EMG signal was discussed. The instantaneous median frequency of simulated and real surface EMG signals were estimated and compared with the conditional mean frequency by means of the methods of time frequency analysis including smoothed pseudo Wigner Ville distribution, the Choi Williams distribution, spectrograms and scalograms. Advantages and drawbacks of the involved methods were compared. Result Both the instantaneous median frequency and conditional frequency could track the change of the frequency components of the simulated EMG signal, but the former was smoother than the latter and had smaller relative errors. Experiments with real surface EMG signal showed that the instantaneous median frequency changes more apparently than the conditional frequency during the process of sustained muscle contractions. Further more, the performance of estimation methods aforementioned could be improved by choosing an appropriate type of window and its width. Conclusion The instantaneous median frequency could indicate the change of the frequency component of surface EMG signal more accurately and sensitively than the conditional mean frequency.