Sequential Blind Signal Separation from Source Mixtures

Recently, blind signal separation (BSS) has been exploited in many applications. As a special class of second order statistics, canonical correlation analysis (CCA) allows to study the correlation between two sets of data. A new framework based on CCA techniques is presented by starting with the conventional method that utilizes a known training signal. Several specific transformations are considered to illustrate the utility of CCA. As a result, A CCA based BSS approach is deduced correspondingly. Someone can select to estimate one or a few source signals, thus saving a lot of computation resource. Computer simulations demonstrate its efficiency and accuracy.