Heart sound analysis using the S transform

Abstract

Central to this study was a pathological sound called opening snap (OS). The first objective was to detect this sound in recordings in which it occurred too close in rime to one of the normal sounds, the second heart sound (S2). The second objective was to differentiate between the OS and another sound, the third heart sound (S3), the timing of which is similar to that of OS. Both OS and S3 occur shortly after S2. Three techniques were used and their performance was compared: The Short Time Fourier Transform (STFT), the S Transform (ST) and the Continuous Wavelet Transform (CWT). These transforms yield time-frequency or time-scale representations of the signal. In addition, the ST and the CWT are multiresolution transforms. The ST proved to be the best in meeting the objectives, mainly because it yielded highly distinct patterns for each kind of sound. Being a relatively new transform, the ST had not been applied to heart sounds in the past.