Using Phase Linearity to Tackle the Permutation Problem in Audio Source Separation

This paper describes a method for solving the permutation problem in blind source separation (BSS) by frequencydomain independent component analysis (FD-ICA). FD-ICA is a well-known method for BSS of convolutive mixtures. However, FD-ICA has a source permutation problem, where estimated source components can become swapped at different frequencies. Many researchers have suggested methods to solve the source permutation problem including using correlation between adjacent frequencies. In this paper, we discuss a new method for solving the permutation problem, based on the linearity of the phase response of the FD-ICA de-mixing matrix. Initial results indicate that our method can provide an almost perfect solution to the permutation problem in an anechoic environment, and better performance than the method based on correlation between adjacent frequencies in an echoic environment.