A Novel Algorithm based on Stockwell Transform for Boundary Detection and Segmentation of Heart Sound Components from PCG signal

Abstract

Segmentation is one of the critical phases involved in the analysis of Phonocardiogram (PCG) signal to detect the heart valve related disorders. A segmentation algorithm provides the location of fundamental heart sound components (S1 and S2) and their duration, which helps to distinguish the murmurs present in the heart sound. In this paper, a novel algorithm based on Stockwell transform is proposed for the segmentation and boundary detection of heart sound components present in the PCG signal. In order to validate, the proposed algorithm is evaluated on eight PCG signals (normal/abnormal) collected from Michigan open source heart sound and murmur database. The result obtained shows that the proposed algorithm is able to detect and segment the heart sound components accurately with an overall accuracy of 97.5% and the time required for execution is 1.025 s in each cardiac cycle.