A Beamforming Approach to Permutation Alignment for Multichannel Frequency-Domain Blind Speech Separation

In this work we explore important connections between blind source separation (BSS) and ideal beamforming. We first compare the performance of a null-steering beamformer against that of a frequency-domain BSS method in a reverberant environment, drawing some interesting conclusions. We then examine the feasibility of using beamformer concepts to resolve permutation inconsistency across frequency, which degrades the performance of BSS methods in a reverberant environment. We also propose a permutation alignment scheme based on information gathered from the microphone array directivity patterns. This technique is novel in the sense that it works satisfactorily even when the directivity patterns exhibit grating lobes, where, in fact, better separation can be achieved in principle. We perform experiments that support the viability of the proposed method under different operating conditions and microphone spacings.