An algorithm for solving permutation problem based on the magnitude information received by microphones in the frequency-domain BSS

Since the permutation ambiguity is a major problem in the frequency-domain blind source separation,this paper presents a new method for solving this permutation ambiguity problem.The new method extracts information from the magnitude attenuation of the signal received by the microphones and performs k-means clustering in order to assign the separated speech signals to different sources.This new approach does not require any specific arrangement or any limitation to the shape of microphone array and works with any number of mixed signals.Experimental results show that the new approach which exploits the magnitude information received by the microphones is able to provide a reliable permutation alignment in most frequency bins simply and effectively.