A clustering approach for solving the spatial aliasing problem in convolutive blind source separation
For the separation of convolutive mixtures, an often used approach is the transformation to the time-frequency domain, where the problem is reduced to multiple instantaneous mixtures. This allows for the employment of well-known ICA algorithms. The drawbacks of this method are the inherent permutation and scaling problems. These ambiguities have to be corrected before a transformation back to the time domain can be carried out. The scaling ambiguity is usually solved using the minimal distortion principle. For the permutation problem, several approaches have been proposed. In this paper we propose a modification of an existing algorithm with the aim of simplifying the depermutation criterion and the corresponding computational effort while maintaining the same performance.