Bispectral analysis of heart sounds measured on the surface of the chest

Bispectral analysis is proposed to investigate nonlinear coupling phenomena of heart sounds. A non-Gaussian AR model is selected for parametric bispectral estimation in analysing several kinds of heart sounds. Bispectral cross-correlation is employed to order determination of the model. Real sounds are utilized to show that different quadratic nonlinear interactions exist in both normal and abnormal sounds. It was found that bispectral analysis has a high immunity to background noise. It is expected that polyspectra may be used as an effective and useful tool for understanding the basic heart vibration mechanism and for improving the automated analysis of these sounds.