Second Heart Sound (S2) Decomposition by Hilbert Vibration Decomposition (HVD) for Affective Signal Modeling and Learning

摘要

This article presents a novel signal decomposition method, Hilbert vibration decomposition (HVD), for analyzing one of the major heart sound components second heart sound (S2) signal for affective signal modeling. In this proposed method, three kinds of simulated S2 signals are generated and the typical one is chosen for decomposition. For HVD method, a FIR filter is designed to separate each of the decomposed components. Finally, performance indicators, including the number of decomposed components, Hilbert spectrum, and spectral centroids, are measured. To evaluate the performance of HVD, the decomposed components are compared with those generated by empirical mode decomposition (EMD) method. The experimental result shows that the number of meaningful decomposed components and frequency resolutions by using HVD method are better than those by using EMD. Such results also reveal the HVD method is superior to the normal EMD method, especially for low frequency narrow band bio-signals such second heart sound, thereby facilitating generating discriminant features for model learning.