High resolution ECG filtering using adaptive Bayesian wavelet shrinkage

Abstract

This paper outlines a Bayesian wavelet shrinkage denoising approach for High Resolution ECG (HRECG) filtering. The authors' proposed filtering method comprises three basic steps: the dyadic Wavelet Transform (WT) computation, the shrinkage of the wavelet coefficients using adaptive Bayesian rules, and the reconstruction of the denoised signal through the inverse WT. An automatic, level-dependent scheme is designed to estimate the shrinkage functions, using a maximum likelihood procedure across the WT coefficients from the ensemble of available beats. The performance evaluation using controlled simulation experiments revealed that the present technique outperforms the wavelet soft and hard-thresholding methods in preserving the high-frequency components of the QRS complex.