| Paper | Code | PESQ-WB | SI-SDR | ESTOI | SIGMOS | DNSMOS | POLQA | ModelName | ReleaseDate |
|---|---|---|---|---|---|---|---|---|---|
| Investigating Training Objectives for Generative Speech Enhancement | ✓ Link | 3.09 | 16.29 | 0.73 | 3.18 | 3.72 | 3.71 | Schrödinger Bridge (PESQ loss) | 2024-09-16 |
| Speech Enhancement and Dereverberation with Diffusion-based Generative Models | ✓ Link | 2.50 | 16.78 | 0.73 | 3.41 | 3.88 | 3.40 | SGMSE+ | 2022-08-11 |
| Hybrid Transformers for Music Source Separation | ✓ Link | 2.37 | 16.92 | 0.71 | 2.87 | 3.66 | 2.97 | Demucs v4 | 2022-11-15 |
| Schrödinger Bridge for Generative Speech Enhancement | 2.33 | 17.85 | 0.73 | 3.44 | 3.83 | 3.46 | Schrödinger Bridge | 2024-07-22 | |
| Conv-TasNet: Surpassing Ideal Time-Frequency Magnitude Masking for Speech Separation | ✓ Link | 2.31 | 16.93 | 0.70 | 2.69 | 3.47 | 2.73 | Conv-TasNet | 2018-09-20 |
| Conditional Diffusion Probabilistic Model for Speech Enhancement | ✓ Link | 1.60 | 8.35 | 0.53 | 2.08 | 2.87 | 1.81 | CDiffuSE | 2022-02-10 |