Noise-assisted trend-filtering of fetal-electrocardiogram signals

Trend-filtering of physiological signals is one of the most challenging tasks, owing to the non-linear and non-stationary nature of the signals. In this regard, trend extraction from noisy fetal heart signals is a possible measure for diagnosis of a fetal pathological condition. The detrended fluctuation analysis (DFA) of non-invasive fetal-electrocardiogram (FECG) is principally influenced by the fetal brain activity, myographic artifacts of both the mother and the fetus. This paper proposes to incorporate an improved trend-filtering technique for non-invasive FECG signals. The method is based on the improved complete ensemble empirical mode decomposition (CEEMD) to analyze and filter out low-frequency trends from complex FECG signals. The CEEMD solves the mode mixing problem by introducing a particular noise in each stage of decomposition. Moreover, an automatic intrinsic mode functions (IMF) selection criteria is incorporated to provide a better spectral separation with a less sifting operation. The simulations performed on FECG signals, collected from standard database show promising results in terms of the reconstruction quality.