Schrödinger Bridge (PESQ loss)

Reported on 6 benchmarks across 1 task · 1 paper · 2 SOTA

Note: results are matched by exact model name. Different papers may use the same name for different model variants.

Audio6 results

Speech EnhancementonEARS-WHAM
PESQ-WB· 2024-09-16
3.09
SOTA
Investigating Training Objectives for Generative Speech Enhancement arXiv:2409.10753
Speech EnhancementonEARS-WHAM
POLQA· 2024-09-16
3.71
SOTA
Investigating Training Objectives for Generative Speech Enhancement arXiv:2409.10753
Speech EnhancementonEARS-WHAM
DNSMOS· 2024-09-16
3.72
best: 3.88 (SGMSE+)
Investigating Training Objectives for Generative Speech Enhancement arXiv:2409.10753
Speech EnhancementonEARS-WHAM
ESTOI· 2024-09-16
0.73
Investigating Training Objectives for Generative Speech Enhancement arXiv:2409.10753
Speech EnhancementonEARS-WHAM
SI-SDR· 2024-09-16
16.29
best: 17.85 (Schrödinger Bridge)
Investigating Training Objectives for Generative Speech Enhancement arXiv:2409.10753
Speech EnhancementonEARS-WHAM
SIGMOS· 2024-09-16
3.18
best: 3.44 (Schrödinger Bridge)
Investigating Training Objectives for Generative Speech Enhancement arXiv:2409.10753