A Canonical Correlation Approach to Blind Source Separation

A method based on canonical correlation analysis (CCA) for solving the blind source (BSS) problem is presented. In contrast to independent components analysis (ICA), the proposed method utilises the autocorrelation in the source signals. This makes the BSS problem easier to solve than if only the statistical distribution of the sample values is considered. Experiments show that the method is much more computationally efficient than ICA. The proposed method can be seen as a generalization of the maximum autocorrelation factors (MAF) method.