Speech Separation on WSJ0-4mix (test)
12.9SI-SNRiVoice Separation
Evaluation results
| Method | Links | |
|---|---|---|
| Voice SeparationTraining Data=WSJ0-Nmix, Training with variable number of speakers=false, Architecture=multiple2020.06 | 12.9 | |
| MULCAT#params=7.5M2020.02 | 12.9 | |
| Conditional TasNet*Training Data=WSJ0 2-5 mix, Training with variable number of speakers=true, Architecture=single2020.06 | 12.5 | |
| IBMtraining=from original paper2020.02 | 10.6 | |
| DPRNNTraining Data=WSJ0-Nmix, Training with variable number of speakers=false, Architecture=multiple2020.06 | 10.4 | |
| DPRNN#params=3.6M, training=by paper authors2020.02 | 10.4 | |
| OR-PITTraining Data=WSJ0-2&3mix, Training with variable number of speakers=true, Architecture=single2020.06 | 10.2 | |
| IRMtraining=from original paper2020.02 | 9.8 | |
| ConvTasNetTraining Data=WSJ0-Nmix, Training with variable number of speakers=false, Architecture=multiple2020.06 | 8.5 | |
| ConvTasNet#params=5.1M, training=by paper authors2020.02 | 8.5 |