Multi-channel speech separation on Librispeech + DEMAND (out-of-domain)
2.15PESQSpatialNet
Evaluation Results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| SpatialNetNFE=1, RTF=0.0078, Type=D2026.06 | 2.15 | 0.8 | 10 | 2.87 | 3.15 | 3.76 | |
| SpatialNet + Fast-GeCoNFE=1+1, RTF=0.0078+0.0068, Type=D+G2026.06 | 2.13 | 0.83 | 10.18 | 3.09 | 3.54 | 4.09 | |
| SpatialNet + MeCoNFE=1+1, RTF=0.0078+0.0068, Type=D+G2026.06 | 2.13 | 0.84 | 10.41 | 3.22 | 3.75 | 4.52 | |
| CrossNetNFE=1, RTF=0.0065, Type=D2026.06 | 2.05 | 0.77 | 8.96 | 2.82 | 3 | 3.68 | |
| CrossNet + Fast-GeCoNFE=1+1, RTF=0.0065+0.0068, Type=D+G2026.06 | 2.02 | 0.8 | 9.18 | 3.05 | 3.41 | 4 | |
| CrossNet + MeCoNFE=1+1, RTF=0.0065+0.0068, Type=D+G2026.06 | 2.02 | 0.82 | 9.38 | 3.16 | 3.62 | 4.45 | |
| SpatialNet + MeanFlowNFE=1+1, RTF=0.0078+0.0068, Type=D+G2026.06 | 1.98 | 0.82 | 10.43 | 3.07 | 3.65 | 4.41 | |
| CrossNet + MeanFlowNFE=1+1, RTF=0.0065+0.0068, Type=D+G2026.06 | 1.89 | 0.8 | 9.39 | 3.03 | 3.56 | 4.35 | |
| DeFTAN2NFE=1, RTF=0.0155, Type=D2026.06 | 1.78 | 0.71 | 4.96 | 2.88 | 2.9 | 3.62 | |
| DeFTAN2 + Fast-GeCoNFE=1+1, RTF=0.0155+0.0068, Type=D+G2026.06 | 1.75 | 0.72 | 5.1 | 3.08 | 3.22 | 4 | |
| DeFTAN2 + MeCoNFE=1+1, RTF=0.0155+0.0068, Type=D+G2026.06 | 1.75 | 0.73 | 5.19 | 3.17 | 3.41 | 4.38 | |
| DeFTAN2 + MeanFlowNFE=1+1, RTF=0.0155+0.0068, Type=D+G2026.06 | 1.66 | 0.71 | 5.18 | 3.04 | 3.34 | 4.26 |