Feature extraction and recognition of heart sound

A novel heart sound (HS) recognition system based Full Bayesian Neural Network Model (FBNNM) is presented. Features are extracted from Power Spectrum Distribution (PSD), the normalized average Shannon energy (NASN) in wavelet domain; and the Correlation-Dimension (CD) is also employed from the view of dynamic system. These features were used as inputs to the FBNNM for HS recognition. A total of 64 samples are used to train FBNNM for 8 different types of HS. The results show the suggested methodology offers a promising recognition of HS.