Optimum denoising method of the uterine EHG signal

The uterine electromyogram (EMG) or electrohysterogram (EHG) is an electrophysiological signal which represents a woman´s uterine activity during pregnancy. This signal is corrupted by electronic and electromagnetic noise and electrophysiological signals. Wavelet theory is very well suited to this nonstationary signal. The use of an undecimated wavelet transform for denoising has already given very satisfying results. The representation of the noise corrupting the EHG signal in adapted frequency bands (undecimated wavelet packet transform) allowed us to discover that this noise was principally composed of the mother´s electrocardiogram (ECG) signal. This ECG noise remains in spite of analogical filtering during acquisition. In this paper, we propose an improvement of our method, based on undecimated wavelet packet shrinkage with coefficient thresholding adapted to the rejection of the maternal ECG