Blind separation for fetal ECG from single mixture by SVD and ICA

Abstract

In this thesis, we propose a novel blind-sourceseparation method to extract fetal ECG from a single-channelsignal measured on the abdomen of the mother. The signal is amixture of the fetal ECG, the maternal ECG and noise. The key ideais to compute the spectrogram of the original signal, and then usean assumption of statistical independence between the componentsto find the trends of the original signal. This is achieved byapplying Singular Value Decomposition (SVD) on the spectrogram,followed by an iterated application of Independent ComponentAnalysis (ICA) on the principle components. The SVD contributes tothe separability of each component and the ICA contributes to theindependence of the two components. We further refine and adaptthe above general idea to ECG by exploiting a-prior knowledge ofthe maternal ECG frequency distribution and other characteristicsof ECG. Experimental studies show that the proposed method is moreaccurate than using SVD only. Because our method does not exploitextensive domain knowledge of the ECGs, the idea of combining SVDand ICA in this way can be applied to other blind separationproblems.