Application to biomedical signals - Practical Biomedical Signal Analysis Using MATLAB - page 192

and [Khamene and Negahdaripour, 2000] who used biorthogonal quadratic spline
wavelet. Two-level wavelet transform (Daubechie wavelets) followed by low-pass fil-
tering was applied by [Hassanpour and Parsaei, 2006]. [Jafari and Chambers, 2005]
used for fECG extraction sequential source separation in the wavelet domain.
The problem of fECG extraction from the signal registered on the abdomen can
be formulated as a blind source separation problem, since it involves the separation
of the original source signals from a sensor array, without the knowledge on the
transmission channels. BSS method involving singular value decomposition and ICA
was applied to fECG by [De Lathauwer et al., 2000]. It was found that in the job
of separation of fECG from mECG the independent component method performed
better than PCA. The results of ICA application to eight channel data is shown in
Figure 4.40.
FIGURE 4.40: Application of ICA for fECG extraction. Panel a): eight-channel
set of cutaneous data recording (5 upper channels placed on abdomen, 3 lower chan-
nels placed on thorax). Panel b): results of ICA application to the signals shown
in panel a). mECG visible in the three upper traces, fECG best visible in the sixth
signal. From [De Lathauwer et al., 2000].
In solving the problem of fECG - mECG separation, especially by ICA- based
BSS approaches, some theoretical limitations appear, due to the requirement that
the number of observations (registered signals) must be equal to or higher than the
number of uncorrelated signal sources. However, in addition to heart signals, each
electrode also picks up activity due to maternal myoelectric activity and other noises,
which are the sources of partly uncorrelated noise different for each channel. There-
fore BSS requires multielectrode arrays, usually including electrodes placed on tho-
rax, which is not always practical in the clinical environment.
The aim of several works concerning fECG was the extraction of a fetal heart rate
variability (fHRV) signal which can be used to find the synergetic control activity
of the sympathetic and parasympathetic branches of autonomous nervous system