EWT-IIT: a surface electromyography denoising method

Abstract

Surface electromyography (sEMG) is often interfered by noise, which has a very important impact on the follow-up research based on sEMG signals, such as motion intention recognition, disease diagnosis, and human–computer interaction. In this paper, an sEMG denoising algorithm based on empirical wavelet transform (EWT) and improved interval thresholding (IIT) is proposed to eliminate noise interference of sEMG signals. The proposed method uses EWT to decompose the original sEMG with noise into several empirical intrinsic modal functions (EIMFs) and then applies the IIT function proposed in this paper to conduct threshold processing for each EIMF; this method is called EWT-IIT. Ten healthy subjects participated in the experiment; the corresponding sEMG signals were analyzed. The signal-to-noise ratio (SNR), root mean square error (RMSE), and ${{\varvec{R}}}^{2}$ were used to evaluate the effect of denoising. The simulated and experimental results show that the IIT function proposed in this paper combines the advantages of hard threshold function and soft threshold function, and EWT-IIT method can effectively remove the noise with the best denoising effect.

Graphical abstract

Appendix

The function to add white Gaussian noise to signal.

function [Y,NOISE] = noisegen(X,SNR)

% noisegen add white Gaussian noise to a signal.

% [Y, NOISE] = NOISEGEN(X,SNR) adds white Gaussian NOISE to X. The SNR is in dB.

NOISE=randn(size(X));

NOISE=NOISE-mean(NOISE);

signal_power = 1/length(X)*sum(X.*X);

noise_variance = signal_power / ( 10^(SNR/10) );

NOISE=sqrt(noise_variance)/std(NOISE)*NOISE;

Y=X+NOISE;