A fast and adaptive ICA algorithm with its application to fetal electrocardiogram extraction

In many applications, it is often required to separate mixtures of sub-Gaussian, skewed, near Gaussian, and super-Gaussian source signals so as to obtain desired source signals. In this paper, we propose a fast and adaptive ICA algorithm based on fully-multiplicative orthogonal-group (FastAdaptiveOgICA), which not only can instantaneously separate mixtures of sub-Gaussian and super-Gaussian source signals, but also can separate skewed and/or near Gaussian signals, which are common in some key application areas, such as biomedical signal processing and telecommunications. Separation performance highly depends on nonlinear functions, so a self-adaptive nonlinear function, which adjusts its parameters to achieve better performance according to the estimation of moments of source signals, is presented in our algorithm. We successfully applied the algorithm to obtain fetal electrocardiogram (FECG) signals with better separation performance and faster convergence speed, compared with several famous ICA algorithms.