Segmentation of S1–S2 sounds in phonocardiogram records using wavelet energies

Locating S1 and S2 sounds in order to diagnose diseases by classifying through the determination of systole and diastole phases is of great significance. The method suggested in this study is available for the segmentation of S1-S2 sounds in heart sounds (HSs) acquired in real-time and it also renders the classifying possible, thereby ensuring a correct diagnosis. In this study, multi-band wavelet energy (WTE) method was proposed for the segmentation of S1-S2 sounds in 16 different HS types. After normalizing the heart sound, the signal is filtered by using wavelet transform, after than S1-S2 sounds are segmented by using multi-band wavelet energy method. For comparison, besides the method which is proposed, two different methods are investigated. One of them is segmentation of S1-S2 sounds using multi-band wavelet Shannon energy (WSE) method and the other is segmentation of S1-S2 sounds using homomorphic filtering (HMF) method. The highest performances are achieved by the proposed WTE method; 91% and 89% segmentation accuracies are obtained for SI and S2 sounds, respectively. The methods' robustness to noise was also analysed.