A Wavelet Based Solution to Extract AN Components from Electromyogram

Wavelet transform has demonstrated its exception power on digital imaging processing, moreover in recent decades wavelet time-frequency distribution has been successfully applied into the biomedical images decomposition and reconstruction process including analysis and integration of electromyography (EMG). Whereas in medical practice one of the major drawbacks in clinical EMG diagnosis lies in the inefficiency on spiked components identification which have small amplitude but possess great value on Alcoholic Neuropathy (AN) diagnosis. In clinical EMG diagnosis time-frequency components with small amplitude or time transient characters are hard to be figured out owing to EMGpsilas masking effect upon these components and the presence of high-energy slow waves within image reconstruction interval. Aiming at this problem this paper puts forward a wavelet-based algorithm to attenuate EMG's masking effect, meantime weaken impact strength of transient noise interference. Wavelet transform is adopted and integrated into EMG data preprocessing operation, within whose process wavelet approximation is filtered out while wavelet details are extracted for further treatments. Wavelet coefficients treatment principle is based on the kurtosis probability theorem and error minimum square value criterion. Numerical simulation is implemented following above algorithm, whose computation consequence reveals that image characters of small amplitude AN components is strengthened by attenuating EMG's masking effects, meanwhile the valuable original image signatures is retained for latter analysis. Numerical results with and without wavelet preprocessing are listed for comparison, which indicate image readability degree is enhanced obviously, moreover there also exists potentials for further improvements.