Energy predictability to blind source separation

A blind source separation (BSS) method based on the energy (square) predictability of original sources is proposed. The method exploits the nonstationarity of sources in the sense that the variance of each source signal can be assumed to change smoothly against time. In contrast to linear predictability, it is shown that nonlinear predictability can also be used for BSS. Simulations verify the efficient implementation of the proposed method, especially its robustness to the outliers.