Speech Separation on WHAM! (test)
18.1SI-SNRi (dB)MossFormer2
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| MossFormer2Para.(M)=55.72023.12 | 18.1 | — | |
| SepReformer-MParams. (M)=17.3, MACs (G/s)=81.32024.06 | 17.8 | 18.1 | |
| SPMamba(OA)Attention mechanism=Omni-directional Attention (OA)2026.01 | 17.7 | 17.9 | |
| SepReformer-BParams. (M)=14.2, MACs (G/s)=39.82024.06 | 17.6 | 18 | |
| SPMamba2026.01 | 17.4 | 17.6 | |
| SepReformer-SParams. (M)=4.3, MACs (G/s)=21.32024.06 | 17.3 | 17.7 | |
| MossFormerPara.(M)=42.12023.12 | 17.3 | — | |
| SepReformer-TParams. (M)=3.5, MACs (G/s)=10.42024.06 | 17.2 | 17.5 | |
| TF-GridNet(OA)Attention mechanism=Omni-directional Attention (OA)2026.01 | 17.1 | 17.2 | |
| TF-GridNetParameters=14.4M2026.01 | 16.9 | 17.2 | |
| ISCITParams. (M)=58.4, MACs (G/s)=252.2, Additional speaker information=true2024.06 | 16.4 | 16.8 | |
| SepFormerDM=true2022.02 | 16.4 | 16.7 | |
| SepFormerPara.(M)=25.72023.12 | 16.4 | — | |
| SepFormerSpeedA.=true2022.02 | 16.3 | 16.7 | |
| DPTNET-SRSSN2021.10 | 16.1 | 16.5 | |
| WaveSplitParams. (M)=29.02024.06 | 16 | 16.5 | |
| WavesplitData augmentation=true2021.10 | 16 | 16.5 | |
| WavesplitDM=true2022.02 | 16 | 16.5 | |
| WavesplitPara.(M)=292023.12 | 16 | — | |
| DPRNN-SRSSN2021.10 | 15.7 | 16.1 | |
| WavesplitData augmentation=false2021.10 | 15.4 | 15.8 | |
| TDANetParams. (M)=2.3, MACs (G/s)=9.12024.06 | 15.2 | 15.4 | |
| Nachmani et al.2021.10 | 15.2 | — | |
| TDANet LargeParams (M)=2.3, kernel length=2ms, stride size=0.5ms2022.09 | 15.2 | 15.4 | |
| VSUNOSPara.(M)=7.52023.12 | 15.2 | — | |
| TDANet2026.01 | 15.2 | 15.4 | |
| DPTNetParams. (M)=2.7, MACs (G/s)=102.52024.06 | 14.9 | 15.3 | |
| DPTNET-TASNET2021.10 | 14.9 | 15.3 | |
| DPTNetParams (M)=2.72022.09 | 14.9 | 15.3 | |
| TDANetParams (M)=2.32022.09 | 14.8 | 15 | |
| SepformerParams. (M)=26.0, MACs (G/s)=86.92024.06 | 14.7 | 15 | |
| SepFormer2022.02 | 14.7 | 15.1 | |
| A-FRCNNParams. (M)=6.1, MACs (G/s)=125.02024.06 | 14.5 | 14.8 | |
| A-FRCNN-16Params (M)=6.12022.09 | 14.5 | 14.8 | |
| A-FRCNN2026.01 | 14.5 | 14.8 | |
| SepformerParams (M)=26.02022.09 | 14.4 | 15 | |
| DPRNN-TASNET2021.10 | 13.9 | — | |
| DPRNNPara.(M)=2.62023.12 | 13.9 | — | |
| SuDORM-RFParams. (M)=6.4, MACs (G/s)=10.12024.06 | 13.7 | 14.1 | |
| DPRNNParams. (M)=2.6, MACs (G/s)=88.52024.06 | 13.7 | 14.1 | |
| SuDORM-RF 2.5xParams (M)=6.42022.09 | 13.7 | 14.1 | |
| DualPathRNNParams (M)=2.72022.09 | 13.7 | 14.1 | |
| SudoRM-RF2026.01 | 13.7 | 14.1 | |
| DualPathRNN2026.01 | 13.7 | 14.1 | |
| MGST2022.02 | 13.1 | — | |
| Learnable fbank2021.10 | 12.9 | — | |
| Cascaded-Bi-LSTM-TASNET2021.10 | 12.9 | — | |
| SuDORM-RF 1.0xParams (M)=2.72022.09 | 12.9 | 13.3 | |
| Learnable fbank2022.02 | 12.9 | — | |
| Conv-TasNetParams. (M)=5.1, MACs (G/s)=10.52024.06 | 12.7 | — | |
| Conv-TASNET2021.10 | 12.7 | — | |
| Conv-TasNetParams (M)=5.62022.09 | 12.7 | — | |
| Conv-TasNet2022.02 | 12.7 | — | |
| Conv-TasNetPara.(M)=5.12023.12 | 12.7 | — | |
| Conv-TasNet2026.01 | 12.7 | 13 | |
| Bi-LSTM-TASNET2021.10 | 12 | — | |
| Chimera++2021.10 | 9.9 | — | |
| BLSTM-TasNetParams (M)=23.62022.09 | 9.8 | — |