Selective noise filtering of high resolution ECG through wavelet transform

A wavelet transform (WT) based filtering of the high resolution electrocardiogram is proposed in this study. The overcomplete time-scale decomposition of the row signal is achieved in the first step by convolution with the quadratic spline wavelet. The small scale data are passed by the filtering algorithm at locations where the wavelet coefficients correlation at adjacent scales is high and suppressed if the correlation is small. The filtered signal is reconstructed from the set of filtered scales through the inverse WT. A comparison with the Wiener filter proved that the wavelet filtering performed better in passing the useful high-frequency data from the QRS terminal part. The denoising results obtained on real signals were similar with 0.6 /spl mu/V root-mean-square noise endpoint signal averaged ECG.