From: Learning-based robust speaker counting and separation with the aid of spatial coherence
MAC
sG1
G1
G2
G3
1
0.923
0.918
0.906
0.910
0.919