A new algorithm for blind extraction of time-correlated fetal electrocardiogram signals

Non-invasive fetal electrocardiogram (FECG) extraction is a vital issue in biomedical signal processing and analyses. Many algorithms have been proposed, but they all have some faults. An objective function is introduced combining the non-Gaussianity and the autocorrelations of the signal sources. A new algorithm is developed by using an approximative Newton method. Comparison results show that the algorithm is simpler to implement because it does not need any iterative step than the existing gradient algorithm. Simulations illustrate that the new algorithm extracts the clear FECG effectively and is very robust to the estimated error of time-delay.