Speech Enhancement On Ears Wham
Métriques
DNSMOS
ESTOI
PESQ-WB
POLQA
SI-SDR
SIGMOS
Résultats
Résultats de performance de divers modèles sur ce benchmark
Nom du modèle | DNSMOS | ESTOI | PESQ-WB | POLQA | SI-SDR | SIGMOS | Paper Title | Repository |
|---|---|---|---|---|---|---|---|---|
| Demucs v4 | 3.66 | 0.71 | 2.37 | 2.97 | 16.92 | 2.87 | Hybrid Transformers for Music Source Separation | |
| Conv-TasNet | 3.47 | 0.70 | 2.31 | 2.73 | 16.93 | 2.69 | Conv-TasNet: Surpassing Ideal Time-Frequency Magnitude Masking for Speech Separation | |
| SGMSE+ | 3.88 | 0.73 | 2.50 | 3.40 | 16.78 | 3.41 | Speech Enhancement and Dereverberation with Diffusion-based Generative Models | |
| Schrödinger Bridge (PESQ loss) | 3.72 | 0.73 | 3.09 | 3.71 | 16.29 | 3.18 | Investigating Training Objectives for Generative Speech Enhancement | |
| Schrödinger Bridge | 3.83 | 0.73 | 2.33 | 3.46 | 17.85 | 3.44 | Schrödinger Bridge for Generative Speech Enhancement | - |
| CDiffuSE | 2.87 | 0.53 | 1.60 | 1.81 | 8.35 | 2.08 | Conditional Diffusion Probabilistic Model for Speech Enhancement |
0 of 6 row(s) selected.