Speech Restoration on REVERB (SimData)
2.001CDDCN
Evaluation results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| DCNTraining distribution=Prior dereverberation method2025.09 | 2.001 | 0.225 | 13.326 | 2.935 | 5.269 | |
| TCN+SATraining distribution=Prior dereverberation method2025.09 | 2.2 | 0.24 | 13.06 | 2.58 | 5.17 | |
| GCRNTraining distribution=Prior dereverberation method2025.09 | 2.534 | 0.325 | 10.753 | 1.934 | 4.861 | |
| D-MFMVDRTraining distribution=Prior dereverberation method2025.09 | 2.639 | 0.316 | 9.649 | 2.167 | 4.892 | |
| TF-RestormerTraining distribution=OOD restoration (no REVERB-specific training), Evaluation setup=unified setup2025.09 | 2.88 | 0.489 | 8.139 | 2.825 | 8.004 | |
| WRNTraining distribution=Prior dereverberation method2025.09 | 3.59 | 0.47 | 4.8 | — | 3.59 | |
| WPETraining distribution=Prior dereverberation method2025.09 | 3.748 | 0.514 | 4.864 | 1.722 | 4.22 | |
| Input2025.09 | 3.975 | 0.574 | 3.617 | 1.503 | 3.687 |