A new permutation method for ICA (independent component analysis)- based blind separation is presented. The objective here is to concatenate ICA separated signals in adjacent time blocks. The proposed method utilizes the associativity in column vectors of an estimated mixing matrix and a tracking filter. It has advantages of no required array calibration, efficient computation, and real time updatability. These superiorities are highly beneficial for radar or communication system type applications. Furthermore it is shown, by an experiment using radio wave signals, that the proposed permutation algorithm successfully permutes the separated signal blocks.
展开▼
