5-speaker speech separation on WSJ0-5mix (evaluation)
11.14SDRiTasTas
Evaluation Results
| Method | Links | |
|---|---|---|
| TasTasIdentity-aware=true, blocks_stage_1=8, blocks_stage_2=9, online remixing=true2020.09 | 11.14 | |
| Nachmani's2020.09 | 10.5 | |
| TasTasIdentity-aware=true, blocks_stage_1=6, blocks_stage_2=62020.09 | 10.41 | |
| IRMDescription=Ideal Ratio Mask (Upper bound for STFT based methods)2020.09 | 9.6 | |
| DPRNN-TasNet2020.09 | 8.4 |