Discrimination of heart sounds using higher-order statistics

A new way of heart sounds discrimination based on higher-order statistics is presented in this paper. The diagnostic character of cardiac auscultation is examined in a new domain, the higher-order spectra domain. The proposed analysis takes into account the nonlinearities of the signal, since it preserves the phase information, and assumes that heart sounds are non-Gaussian processes. In this way, a nonlinear analysis of heart sounds is introduced expanding the process of heart sounds to more accurate estimations of their true character. The proposed analysis was conducted on preclassified signals selected from a teaching tape. The derived results have shown that higher-order statistics define new differences in heart sounds analysis, characterizing efficiently various cardiac pathologies or the severity of cardiac dysfunction associated with the heart sounds.