Multiscale analysis of heart sound for segmentation using multiscale Hilbert envelope

Abstract

In this paper a heart sound segmentation algorithm in multiresolution domain is proposed. Wavelet decomposition of heart sound signal grossly segments its components into different subbands. If multiscale Hilbert envelope is computed on reconstructed signals at different scales, it provides suitable markers for first and second heart sound boundaries. The selection of wavelet subband for marker generation is based on analysis of heart sound spectra and energy contribution of wavelet subbands. The heart sound boundaries for S1 and S2 are decided by markers derived from second derivative of Hilbert envelope of the reconstructed subband signal. The proposed method is evaluated using heart sound signal available in the web site of the Department of Medicine, Washington University. The performance of the proposed method is found satisfactory.