Speech Separation on Libri2Mix
Metric: SI-SDRi (higher is better)
Metric: SI-SDRi (higher is better)
| # | Model↕ | SI-SDRi▼ | Extra Data | Paper | Date↕ | Code |
|---|---|---|---|---|---|---|
| 1 | MossFormer2 (w speed perturb) | 22.2 | No | - | - | Code |
| 2 | TF-Locoformer (M) | 22.1 | No | TF-Locoformer: Transformer with Local Modeling b... | 2024-08-06 | Code |
| 3 | MossFormer2 (w/o DM) | 21.7 | No | - | - | Code |
| 4 | Separate And Diffuse | 21.5 | No | Separate And Diffuse: Using a Pretrained Diffusi... | 2023-01-25 | - |
| 5 | WHYV | 17.5 | No | Wanna hear your voice? A sample is all we need! | 2024-10-01 | - |
| 6 | TDANet Large | 17.4 | No | An efficient encoder-decoder architecture with t... | 2022-09-30 | Code |
| 7 | TDANet | 16.9 | No | An efficient encoder-decoder architecture with t... | 2022-09-30 | Code |
| 8 | Conv-Tasnet (Libri1Mix speech enhancement pre-trained) | 14.1 | Yes | Stabilizing Label Assignment for Speech Separati... | 2020-10-29 | Code |
| 9 | Conv-Tasnet (Libri1Mix speech enhancement multi-task) | 13.7 | Yes | Stabilizing Label Assignment for Speech Separati... | 2020-10-29 | Code |
| 10 | Conv-Tasnet | 13.2 | No | Stabilizing Label Assignment for Speech Separati... | 2020-10-29 | Code |