EMG mean power frequency determination using wavelet analysis

Abstract

Surface electromyography (EMG) is routinely used to characterize muscle activity and fatigue in many physiological and pathological circumstances. Commonly-used parameters describing the spectral content of EMG signals include Fourier-based mean and median power frequencies. The application of the Fourier transform to a data stream assumes wide-sense stationarity of the data over the sampled window. In cases where repetitive or non-constant muscular activity is considered, stationarity constraints are violated. The authors propose the use of multiresolution wavelet analysis, which provides both temporal and frequency resolution of a signal, as an alternative method by which to determine the mean power frequency (MPF) of signals with rapidly-varying frequency content. In order to determine the ability of this method to accurately predict MPF, Fourier- and wavelet-derived analyses of isometric, constant-force contractions are compared. Fourier and wavelet-derived MPF calculations from EMG of the large hand flexor muscle group recorded during 20-second isometric contractions at 40% maximum voluntary contraction were compared. The results indicate that the wavelet analysis method consistently over-estimates the Fourier-derived MPF, but that the signal characteristics and shapes of MPF curves are similar.