A novel cardiac spectral segmentation based on a multi-Gaussian fitting method for regurgitation murmur identification

A new cardiac spectral segmentation method was developed for discriminating between normal heart sound and heart valvular diseases. This approach was based on a multi-Gaussian fitting algorithm of cardiac spectral curve. The spectral autoregressive power spectral density (aPSD) curve was estimated from the cardiac sounds noise-cancelled by the wavelet decomposition. 5-GaPSD was approximated by a five-Gaussian model consisting of five Gaussian peaks, P1 to P5. The spectral profiles, the maximum frequency f(k), the amplitude H-k, the half-width W-k, the area portion S-k, and the loss of area, of five Gaussian peaks were investigated and compared for segmenting the spectral information of normal heart sound and two regurgitation murmurs. (C) 2014 Elsevier B.V. All rights reserved.