Spectral analysis of Carpentier-Edwards prosthetic heart valve sounds in the aortic position using SVD-based methods

The authors examine the performance of spectral estimation techniques when analysing the spectral content of Carpentier-Edwards (C-E) prosthetic heart valve sounds. The performance of the following methods is analysed: the fast Fourier transform (FFT), Burg algorithm (BA), sinusoidal signal identification, several algorithms for autoregressive moving average modelling, and Prony´s spectral method (PSM). Synthesised second heart sounds were generated to allow a vigorous comparison between all the methods. From the results achieved, it is concluded that parametric methods based on singular value decomposition are generally more appropriate than the FFT and BA algorithm. Amongst these parametric methods, PSM achieves the best results. Results also show that spectral analysis of closing sounds produced by C-E valve in the aortic position represents a useful technique for monitoring the condition of the prosthetic heart valves