Speech Enhancement on VCTK-DEMAND
3.2PESQWavLM-Large (log 1p + PF)
Evaluation results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| WavLM-Large (log 1p + PF)Backbone=WavLM-Large, Fine-tuning Strategy=log 1p + PF2022.04 | 3.2 | 4.53 | 3.6 | 3.88 | 0.957 | |
| WavLM-Large (EF)Backbone=WavLM-Large, Fine-tuning Strategy=EF2022.04 | 3.17 | 4.49 | 3.58 | 3.85 | 0.956 | |
| WavLM-Base (log 1p + PF)Backbone=WavLM-Base, Fine-tuning Strategy=log 1p + PF2022.04 | 3.16 | 4.5 | 3.57 | 3.86 | 0.956 | |
| PFPL2022.04 | 3.15 | 4.18 | 3.6 | 3.67 | 0.95 | |
| WavLM-Large (log 1p + EF)Backbone=WavLM-Large, Fine-tuning Strategy=log 1p + EF2022.04 | 3.15 | 4.5 | 3.59 | 3.85 | 0.957 | |
| WavLM-Large (PF)Backbone=WavLM-Large, Fine-tuning Strategy=PF2022.04 | 3.14 | 4.47 | 3.57 | 3.82 | 0.957 | |
| SSPF2022.04 | 3.13 | 4.3 | 3.61 | 3.72 | 0.95 | |
| WavLM-Base (log 1p + EF)Backbone=WavLM-Base, Fine-tuning Strategy=log 1p + EF2022.04 | 3.12 | 4.49 | 3.56 | 3.83 | 0.956 | |
| WavLM-Base (EF)Backbone=WavLM-Base, Fine-tuning Strategy=EF2022.04 | 3.11 | 4.44 | 3.56 | 3.79 | 0.955 | |
| WavLM-Base (PF)Backbone=WavLM-Base, Fine-tuning Strategy=PF2022.04 | 3.09 | 4.42 | 3.54 | 3.77 | 0.955 | |
| WavLM-Large (log 1p)Backbone=WavLM-Large, Fine-tuning Strategy=log 1p2022.04 | 3.09 | 4.45 | 3.53 | 3.79 | 0.954 | |
| WavLM-Base (log 1p)Backbone=WavLM-Base, Fine-tuning Strategy=log 1p2022.04 | 3.05 | 4.4 | 3.52 | 3.74 | 0.952 | |
| PERL2022.04 | 3.04 | 4.23 | 3.42 | 3.63 | 0.947 | |
| WavLM-Large (TFS)Backbone=WavLM-Large, Fine-tuning Strategy=Training From Scratch (TFS)2022.04 | 2.87 | 4.24 | 3.41 | 3.57 | 0.945 | |
| WavLM-Base (TFS)Backbone=WavLM-Base, Fine-tuning Strategy=Training From Scratch (TFS)2022.04 | 2.83 | 4.21 | 3.41 | 3.53 | 0.946 | |
| Huang2022.04 | 2.8 | — | — | — | 0.945 |