Classification Of Heart Sounds Using Time-Frequency Analysis And Image Processing Methods

Abstract

This paper describes the algorithm for detection of heart sounds from phonocardiogram record (PCG) using time-frequency analysis and image processing technique. Presented algorithm extracts time-frequency features using continuous wavelet transform (CWT) and subsequently applies image processing methods such as morphological operations and connected components labeling on hearth sounds segment S1-S2. Furthermore, the segmented PCG signal undergoes correction procedure, which uses the Smith-Waterman local alignment algorithm to correctly predict heart sound sequence.