An Efficient Method for Fetal Heart Sounds Detection Based on Hilbert Transform

Abstract

Fetal heart rate (fHR) and fetal heart rate variability (fHRV) are the main biomedical parameters that are used to investigate cardiac disorders and to identify episodes of stress which are exercised on the fetus during pregnancy. Thus, stages of hypoxia can be identified and precise actions can be taken by the physicians. The fetal phonocardiogram (fPCG), which is the representation of the sounds produced by the fetal heart during a cardiac cycle, can be used to derive the fetal heart rate. It has the advantage that it is obtained via a very simple and cost-effective method. However, the main limitation is the very low signal to noise ratio (SNR) because the acoustic sensor records also other events: maternal organ sounds (mOS), maternal heart sounds (mHS) and other acoustic events produced by different sources (background noise, reverberation noise etc). The separation of the fetal heart sounds (fHS) from this acoustic mixture is not simple because some components present high correlation in frequency domain with the fHS. Thus, the most disturbing component is the mHS signal which is narrowband and non-stationary. In this paper is proposed a method for fHS enhancement based on Wavelet and Hilbert transform analysis. The performance of the proposed method in fHS extraction is evaluated with the simulated fPCG database from PhysioBank. Results indicate promising performance in correct localization of the fHS reaching an overall performance of 90%.