Spectral analysis of closing sounds produced by lonescu-Shiley bioprosthetic aortic heart valves

The selection of the optimal number of poles (P) and zeros (Q) for parametric spectral analysis of closing sounds produced by lonescu-Shiley bioprosthetic aortic heart valves was evaluated in 15 patients. These values were obtained by computing the normalised root-mean-square error (NRMSE) function between the aortic closing sounds and the impulse response of the parametric models for different values of P and Q. As expected, the NRMSE function decreased with increasing value of P and Q. The optimal P and Q were selected at the beginning of the plateau of the NRMSE function. For all-pole modelling with autocorrelation and all-pole modelling with covariance, the optimal number of poles is 16. For pole-zero modelling using the Steiglitz-McBride method with maximum entropy and extrapolation to zero, the optimum values are 14 poles and 14 zeros.