DENOISING OF THE UTERINE EHG BY AN UNDECIMATED WAVELET TRANSFORM

We propose two original methods of denoising of the uterine electrohys terography (EHG) signal by wavelets. This external electrophysiologica l signal is corrupted by electronic, electromagnetic noises and by the remaining electrocardiogram of the mother. The interfering signals ha ve overlapping spectra. Therefore, a classical filtering is unusable. Wavelets should be a very well-suited denoising tool. The first propos ed method uses the algorithm ''a trou'' with nonsymmetrical filters. T he computation is rapid and the results are satisfying compared to the classical denoising techniques. The second algorithm is an improvemen t of the first method. It uses orthogonal wavelets and the result of t he thresholding corresponds to the average of all circulant shifts den oised by a decimated wavelet transform. Results are compared to tradit ional denoising algorithms by wavelet (orthogonal, maximally decimated ). The proposed algorithms are more efficient on simulated signals as well as on uterine EHG.