Time-varying Wiener filtering of the fetal ECG using the wavelet transform

The authors discuss decreasing the number of complexes that are used to form an average and then applying a digital filter to this average to obtain a SNR comparable to that when a larger number of complexes are incorporated into the average. They describe the design of such a filter, based upon the Wiener filter and the wavelet transform, capable of reducing the amount of noise present on the averaged fetal ECG (FECG) complex. In the work described, this filter is applied to antenatal data recorded from the maternal abdomen (with the maternal ECG already subtracted). It is shown that this approach can reduce the number of complexes required to produce an acceptable complex by approximately a factor of three. Similar work has already been carried out in the field of evoked potential analysis, but not, to the authors´ knowledge, in the analysis of FECG signals. These initial results suggest the discrete wavelet transform based time-varying Wiener filter will prove of value in reducing the number of complexes that must be time averaged before an FECG complex of sufficient SNR is produced from which timing and amplitude information can be extracted