Classification of native heart valve sounds using the Choi-Williams time-frequency distribution

Describes the novel application of time-frequency (T-F) distributions to the assessment of heart valve condition. The Choi-Williams distribution (CWD) was applied to a set of 45 normal/abnormal valve subjects and 9 descriptive features were extracted from the resulting T-F plots. A definition of normality, for each of the 9 features, was produced and assessing the performance of the 45 data subjects with respect to these criteria resulted in correct classification of 44 of the 45 subjects. This work empirically demonstrates the suitability of T-F techniques as an aid to the assessment of cardiac condition and clearly shows the advantages of T-F over classical spectral estimation techniques.