An Ensemble Empirical Mode Decomposition Based Method for Fetal Phonocardiogram Enhancement

Nowadays, fetal monitoring standard relies mainly on the analysis of fetal heart rate. However, signals like fetal electrocadiogram (fECG) and fetal phonocardiogram (fPCG) can offer complementary diagnostic information derived from the waveform analysis. The limitations of using, in particular, fPCG are: the signal to noise ratio (SNR) is very low because the recorded signal is a mixture of acoustic components originating not only from the fetus heart but also from the mother (maternal heart sounds (MHS), maternal organ sounds (MOS)) and other sources (power line interference, reverbaration noise, sensor and background noise). Moreover, it is dependent on gestational age, fetal and maternal positions, the data acquisition location. From the noise components the MHS presents a high correlation in the frequency domain with the fetal heart sounds (FHS). Thus, separation of MHS from acoustic recordings is not straightforward. In addition the MHS is a narrowband non-stationary signal. Thus, in this paper is proposed a method for fPCG enhancement from the recorded acoustic mixture based on the Esemeble Empirical Mode Decomposition (EEMD). This approach allows to analyze heart sounds into Intrinsic Mode Functions (IMFs) and it is adaptive and data driven. The performance of the proposed method is evaluated on a database with simulated fPCG signals.