Geometrically Constrained Source Extraction and Dereverberation Based on Joint Optimization

Source extraction, which aims at extracting the target source signals from the observed reverberant mixtures, plays an important role in voice communication and human-machine interfaces. Among the numerous source extraction methods that have been developed, the geometrically constrained (GC) one, which incorporates the direction-of-arrival (DOA) information of the target signals, has demonstrated great potential. However, this method generally suffers from significant performance degradation in strong reverberant environments since it is challenging to obtain in such environments accurate DOA estimates that are needed by the algorithm. To address this problem, we present in this work an iterative algorithm, which integrates the source-wise weighted prediction error (WPE)-based dereverberation principle with the geometrically constrained source extraction method. We show that this algorithm is able to improve the DOA estimation accuracy as well as the source extraction performance.