Speech Enhancement on StillSonicSet Reverb-only
2.011PESQSGMSE
Evaluation Results
| Method | Links | |||||||
|---|---|---|---|---|---|---|---|---|
| SGMSEModel Type=Diffusion2026.03 | 2.011 | 0.632 | 4.031 | 3.166 | 3.882 | 2.775 | 0.943 | |
| StoRMModel Type=Diffusion2026.03 | 1.711 | 0.561 | 4.353 | 3.018 | 3.909 | 2.626 | 0.922 | |
| Dit-FlowModel Type=Flow Matching2026.03 | 1.599 | 0.578 | 3.979 | 3.24 | 3.826 | 2.851 | 0.935 | |
| Noisy (lower bound)Model Type=–2026.03 | 1.324 | 0.497 | 3.383 | 2.08 | 2.386 | 1.684 | 0.895 |