TABLE V.
STOI Comparison Between Speech Separation Methods With Target Dependent Training on IEEE-TIMIT Corpus
| −12 dB | −9 dB | −6 dB | −3 dB | O dB | 3 dB | 6 dB | |
|---|---|---|---|---|---|---|---|
| Noisy | 48.9 | 54.9 | 61.4 | 68.0 | 74.5 | 80.4 | 85.5 |
| DNN+Map | 72.6 | 77.0 | 80.9 | 84.5 | 87.7 | 90.4 | 92.5 |
| DNN+IRM | 71.2 | 76.6 | 81.2 | 85.2 | 88.6 | 91.5 | 93.7 |
| MCA+IRM | 73.4 | 78.3 | 82.6 | 86.3 | 89.6 | 92.2 | 94.2 |
| MCS+IRM | 75.1 | 80.2 | 84.4 | 87.9 | 90.8 | 93.1 | 94.8 |
| DNN+SA | 75.0 | 79.4 | 83.0 | 86.3 | 89.1 | 91.5 | 93.5 |
| MCA+SA | 76.5 | 80.8 | 84.3 | 87.4 | 90.0 | 92.3 | 94.0 |
| MCS+SA | 76.9 | 81.4 | 85.1 | 88.1 | 90.7 | 92.9 | 94.6 |
| MCA+IRM+SA | 76.9 | 81.1 | 84.7 | 87.8 | 90.6 | 92.9 | 94.6 |
| MCS+IRM+SA | 77.4 | 81.9 | 85.5 | 88.5 | 91.1 | 93.3 | 95.0 |