Speech Enhancement on LibriCSS real-recorded
2.963SIGStoRM
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| StoRMTrained dataset=StillSonicSet, RTF=0.4942026.03 | 2.963 | 3.946 | 2.49 | 0.953 | |
| StoRMTrained dataset=WSJ0+Reverb [13], RTF=0.4942026.03 | 2.961 | 3.822 | 2.485 | 0.955 | |
| SGMSETrained dataset=StillSonicSet, RTF=0.5652026.03 | 2.955 | 3.901 | 2.479 | 0.951 | |
| DiT-FlowTrained dataset=WSJ0+Reverb [13], RTF=0.2302026.03 | 2.938 | 3.902 | 2.476 | 0.917 | |
| DiT-FlowTrained dataset=StillSonicSet, RTF=0.2302026.03 | 2.935 | 3.95 | 2.503 | 0.928 | |
| SGMSETrained dataset=WSJ0+Reverb [13], RTF=0.5652026.03 | 2.854 | 3.83 | 2.363 | 0.952 | |
| Noisy (lower bound)Trained dataset=N/A, RTF=N/A2026.03 | 2.756 | 3.521 | 2.209 | 0.895 |