A Comparative Phonocardiography Study: Two Wavelet Based Methods for Fetal Heart Sound Detection

Abstract

The Fetal Heart Rate (FHR) periodic monitoring is an essential part during pregnancy. Auscultation is a noninvasive low cost technique for the heart sound examination. However, Fetal Phonocardiogram (fPCG) signals are correlated with high background noise due to the maternal organs. The aim of this study is to develop an automatic analysis technique to identify the location of the FHSs in fPCGs signals and calculate the FHR. In this paper two denoising methods are proposed and compared in real and simulated fPCGs signals. The signals were analyzed in the Wavelet domain and then two methods were proposed for the detection of the FHSs. The first approach combines the Wavelet Transform with the Fractal Dimension analysis (WT-FD) while the second approach combines the Wavelet Transform with Simplicity Filtering (WT-S), which is computed by the eigenvalue spectrum method. Both methods present good performance with the WT-FD presenting 88% precision, 93% recall and WT-S 71% and 72% respectively in fetal heart sounds detection in simulated signals from -4.4 to -26.7dB, while the WT-S method is more unstable in cases with robust noise than the WT-FD. For the real data preliminary results are presenting and analyzed.