Detection of Heart Sound using Logistic Function Amplitude Moderator and Teager-Kaiser Energy Operator

In this paper, we have proposed a new morphological feature to improve the detection of S1 and S2 sounds from noisy or pathological phonocardiogram (PCG) signals. This feature is based on logistic function amplitude moderator (LFAM) and Teager-Kaiser energy operator (TKEO). The LFAM is used to enhance the intensity of weak heart sounds so that the intensities of envelope peaks are uniform. The LFAM is operated at two modes: low amplitude mode and high amplitude mode. The selection of the operational mode is decided based on probability density function (PDF) derived from intensity based histogram analysis. Then the homomorphic envelope of the resulting waveform is calculated. Finally, TKEO is applied to the envelope signal to track the instantaneous energy of signal oscillation. This feature will capture prominent peaks for heart sounds and nullify the effect of noises and murmurs that have more or less uniform envelope intensity distribution. PCG database from the University of Washington has been used for evaluation. The performance of the proposed method is tested using logistic regression-hidden semi-Markov model (LR-HSMM) classifier and obtain an average F
₁
score of 79.08±3.93.