The electrohysterogram (EHG) signal is mainly corrupted by the mother's ele
ctrocardiogram (ECG), which remains present despite analog filtering during
acquisition. Wavelets are a powerful denoising tool and have already prove
d their efficiency on the EHG, In this paper, we propose a new method that
employs the redundant wavelet packet transform. We first study wavelet pack
et coefficient histograms and propose an algorithm to automatically detect
the histogram mode number. Using a new criterion, we compute a best basis a
dapted to the denoising, After EHG wavelet packet coefficient thresholding
in the selected basis, the inverse transform is applied. The ECG seems to b
e very efficiently removed.