Foetal ECG extraction using blind source separation methods

Three methods to achieve Blind Source Separation are applied to the foetal electrocardiogram (ECG) extraction problem: Principal Component Analysis (PCA), Higher-Order Singular Value Decomposition (HOSVD) and Higher-Order Eigen Value Decomposition (HOEVD). The first one gives uncorrelated source signals by means of second-order tools, while the last two resort to higher-order statistics of the data signals, so higher-order independence is attained. When tested on real ECG data, the last two produce better results than the former, with the HOEVD yielding the best performance, as expected from the theoretical unfolding.