The use of wavelet packets to improve the detection of cardiac sounds from the fetal phonocardiogram

This work evaluates the influence of preprocessing to defect time references of the first and second cardiac sounds (S1 and S2). Four fetal phonocardiogram (PDGf) signals with different signal to noise ratio (SNR) were simulated (150 cycles, 50-400 Hz) using acquired cardiac sounds. The signals were decomposed by Mutual Wavelet Packets and the best tree was de-noised before the signal reconstruction. The PCGf envelope was obtained by Hilbert Transform (HT), and the starting (S), maximum (M), and ending (E) points were detected for S1 and S2. The results were compared without and with the preprocessing stage, finding that the proposed method improves the SNR and the detection of the heart sounds.