Automatic Speech Recognition on ATC Corpus
3.5CER (DS2)ROSE
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| ROSECausal=false, Domain=T, #Param. (M)=36.98, GMAC/S=4.732023.12 | 3.5 | 2.42 | |
| ROSE2023.12 | 3.5 | — | |
| DEMUCSCausal=false, Domain=T, #Param. (M)=68.87, GMAC/S=38.122023.12 | 4.28 | 2.72 | |
| SE-ConformerCausal=false, Domain=T, #Param. (M)=75.28, GMAC/S=44.962023.12 | 4.38 | 2.74 | |
| Noisy2023.12 | 4.66 | 2.81 | |
| Noisy2023.12 | 4.66 | — | |
| TF-GridNetCausal=false, Domain=T-F, #Param. (M)=8.2, GMAC/S=19.22023.12 | 4.87 | 2.83 | |
| DEMUCSCausal=true, Domain=T, #Param. (M)=53.52, GMAC/S=37.832023.12 | 4.89 | 2.75 | |
| DB-AIATCausal=false, Domain=T-F, #Param. (M)=2.81, GMAC/S=41.82023.12 | 4.9 | 2.79 | |
| RVAE-EM-SMethod Type=Dereverberation2023.12 | 5.11 | — | |
| FullSubNet+Causal=false, Domain=T-F, #Param. (M)=8.67, GMAC/S=48.572023.12 | 5.35 | 2.76 | |
| BSRNNCausal=true, Domain=T-F, #Param. (M)=2.96, GMAC/S=2.572023.12 | 5.66 | 2.86 | |
| WD-TCNMethod Type=Dereverberation2023.12 | 6.09 | — | |
| CRNCausal=true, Domain=T-F, #Param. (M)=1.7, GMAC/S=8.82023.12 | 7.97 | 2.91 | |
| MetricGAN+Causal=false, Domain=T-F, #Param. (M)=1.86, GMAC/S=28.512023.12 | 7.98 | 2.89 | |
| MetricGANCausal=false, Domain=T-F, #Param. (M)=1.86, GMAC/S=28.512023.12 | 8.31 | 3.01 | |
| DCCRN-ECausal=true, Domain=T-F, #Param. (M)=3.67, GMAC/S=33.482023.12 | 8.46 | 3.97 | |
| Spec-SubCausal=false, Domain=T-F2023.12 | 9.44 | 6.95 | |
| SGMSECausal=false, Domain=T-F, #Param. (M)=3.89, GMAC/S=152.772023.12 | 9.44 | 4.79 | |
| SEGANCausal=false, Domain=T, #Param. (M)=65.82, GMAC/S=13.392023.12 | 9.74 | 4.31 | |
| CDiffuSECausal=false, Domain=T-F, #Param. (M)=4.28, GMAC/S=190.232023.12 | 10.45 | 5.02 | |
| NKFMethod Type=Acoustic Echo Cancellation (AEC)2023.12 | 10.52 | — | |
| WienerCausal=false, Domain=T-F2023.12 | 10.62 | 7.12 | |
| Align-CRUSEMethod Type=Acoustic Echo Cancellation (AEC)2023.12 | 10.99 | — | |
| DiffuSECausal=false, Domain=T-F, #Param. (M)=3.23, GMAC/S=210.312023.12 | 11.54 | 5.55 | |
| WPEMethod Type=Dereverberation2023.12 | 11.67 | — | |
| NLMSMethod Type=Acoustic Echo Cancellation (AEC)2023.12 | 12.89 | — |