Audio-visual speech separation and enhancement on Mandarin dataset (test)
2.51PESQVISUALVOICE
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| VISUALVOICETraining Data=VoxCeleb2, Model Type=Speaker-independent2021.01 | 2.51 | 0.75 | 6.69 | |
| Ephrat et al.Training Data=AV-Speech, Model Type=Speaker-independent2021.01 | 2.5 | 0.71 | 6.1 | |
| Hou et al.Model Type=Speaker-dependent2021.01 | 2.42 | 0.66 | 2.8 | |
| Gabbay et al.Model Type=Speaker-dependent2021.01 | 2.25 | — | — |