Thank you for your interesting for extraction and analysis of fetal ECG.

. See https://www.jstage.jst.go.jp/article/ihj/50/2/50_2_161/_article or attached PDF.

In brief. We used the novel prototype SAVP-ECG system, with a high resolution (0.076 μV/bits) and sampling rate (2000 Hz/s) (Fig. 1). A modified X,Y,Z-leads system was used to obtain fetal ECGs (Fig. 2A). Electrodes were placed on the mothers at the right (+) and left (-) lower points of the abdomen for X instruction, right lower point (+) and left higher point (-) of the abdomen for Y instruction, and at a zone where fetal heart sounds could be distinctly recognized for Z instruction. Fetal ECGs were recorded for five minutes by this modified X,Y,Z-leads system. We used a silver-silver argentic chloride magnet electrode (TE-18 series, Fukuda Denshi Co. Ltd., Tokyo, Japan) for fetal ECG measurements and an ECG electrode (Red Dot, 3M Health Care, USA) for neonatal ECG measurements.

Noise was removed from the original signals via low-level (0.5 Hz) and high-level (25 Hz) removal filters. We made a template of the QRS signal of the mother, extracted the fetal signal, and removed the maternal QRS wave. We then added 300-500 heart-beats by triggering the fetal QRS complex.

Check out PhysioNet's 2013 challenge on Fetal ECG (http://www.physionet.org/challenge/2013/). The resources at the bottom of the challenge provides some good references, including some review articles.

What about extraction and analysis of fetal MRI?

Methods from nonlinear dynamics are reviewed here: http://arxiv.org/pdf/chao-dyn/9807002.pdf

you can use wavelet transform using daubechies function to extract the feature of the ECG waveform.

you can use extended kalman filter by setting the state equation as dynamic model for generating ECG by using this you can predict maternal ECG and then substract estimated maternal from abdominal then post filter it by using wavelet. I think you can get the fetal ECG.

PhysioNet's 2013 challenge on Fetal ECG provide a good platform to discuss and exchange the methods on it. You could pay attention for its publishment on CINC 2013. I think the paper could be downloaded at the end of this month.

I also attend this challenge. Maybe we could discuss it more details.

You must record the signal from the abdomen of the mother (thus recording the mother ECG together with the fetal ECG) and the mother ECG (from DII, as usually) and subtract statistically the mother ECG signal from the ECG recorded in thye abdomen.