Segmentation of Heart Sound Using Double-Threshold

Abstract

Segmentation of heart sound signal is the first step of analysis and automatic diagnosis of heart sound. Although most existing heart sound segmentation methods perform well in normal signal, they usually have no effect on abnormal signal. In this paper, a method based on double-threshold is proposed for robust segmentation of heart sound. Firstly, the signal is preprocessed by filter based on Hamming windows to eliminate background noises and high-frequency murmurs. Then, the Improved Hilbert-Huang transfom is used to extract the envelope of heart sound. Finally, the envelope is segmented by Double-threshold method and the false segmentation is eliminated according to physiology knowledge. Experimental results show that the proposed method is validated to perform well for both normal and typical abnormal heart sounds.