Eigenvector Algorithms with Reference Signals for Frequency Domain BSS
説明
This paper describes blind source separation (BSS) problems in the frequency domain using an eigenvector algorithm (EVA) with reference signals. The proposed EVA has such an attractive feature that all source signals are separated simultaneously from their mixtures. This is an advantage against the methods using deflation process (e.g., super-exponential method), because those methods sometimes do not work so as to converge to desired solutions, due to deflation failure. Computer simulation demonstrates the validity of the proposed EVA.
