Modelling and processing of phonocardiogram via parametric bispectral approach

Proposes a modern parametric approach to bispectral estimation for modelling and processing of phonocardiograms (PCGs). The procedure is based on an autoregressive model driven by a non-Gaussian white noise. The authors develop a third-order recursion which is employed to the modelling of the heart sound signals. The cumulant-based third-order recursion algorithm is derived to show that the parametric method provide bispectral estimation that are far superior to the conventional estimate in term of bispectral fidelity and its resolution. The parametric method is also used to detect the PCG's phase information which is related with the states of the heart. Moreover, it is insensitive to contamination of the heart sound signals by a general class of noise including additive Gaussian noise. Some results are illustrated and compared.