TABLE IV.

STOI Comparison Between Speech Separation Methods With Multi-SNR Speaker-Pair Dependent Training on SSC Corpus With Insufficient Clean Training Data

	−12 dB	−9 dB	−6 dB	−3 dB	O dB	3 dB	6 dB
Noisy	46.4	52.2	58.8	65.8	72.7	79.2	84.8
DNN+Map	56.0	62.6	68.4	73.1	76.6	79.1	80.8
DNN+IRM	57.0	64.4	71.3	77.2	81.9	85.2	87.5
MCA+IRM	57.6	65.1	72.0	77.9	82.5	85.7	87.8
MCS+IRM	56.7	64.6	71.7	77.6	82.0	84.9	86.7
DNN+SA	63.9	70.8	76.7	81.5	85.3	88.1	90.1
MCA+SA	65.5	72.1	77.8	82.4	85.9	88.4	90.2
MCS+SA	65.8	72.3	77.8	82.2	85.5	87.9	89.6