Application to biomedical signals - Practical Biomedical Signal Analysis Using MATLAB - page 191

0 5 mV-100 mV. When measured on maternal abdomen fECG is additionally dis-

turbed by the muscular activity from abdomen and uterus, since the main frequency
range of fetal and maternal ECG is in the frequency range 0.3-100 Hz, while for the
abdominal muscle it is 10-400 Hz and for uterine contractions 0.01-0.6 Hz. There-
fore the problem of extraction of fECG from other disturbing bioelectric signals is
not a simple one. An example of mECG extraction from abdominal ECG is shown
in Figure 4.39.

.

FIGURE 4.39: Example of abdominal ECG (abdECG) decomposition. Maternal
QRS elimination and smoothing. (a) abdECG signal, (b) the signal after the maternal
QRS elimination (eECG signal). (c) Smoothed eECG signal (seECG). From [Kar-
vounis et al., 2007].

Several methods for extraction of fECG from maternal ECG (mECG) were de-
vised including: adaptive filtering [Martinez et al., 1997], singular value decompo-
sition [Kanjilal et al., 1997], fractals [Richter et al., 1998], neural networks [Camps
et al., 2001], fuzzy logic [Al-Zaben and Al-Smadi, 2006], polynomial networks [As-
saleh and Al-Nashash, 2005], combination of PCA and matched filtering [Martens
et al., 2007], template construction [Taralunga et al., 2008]. Wavelet transform
also found several applications to the fECG analysis, e.g., [Karvounis et al., 2007]