Fig. 15From: Learning-based robust speaker counting and separation with the aid of spatial coherenceComparison of separation performance with array configuration (G1) in terms of a, c PESQ and b, d WER for different overlap ratiosBack to article page