Binaural Target Speech Extraction on WSJ0
3.23DNSMOS OverallBDE-BiTSE
Evaluation results
| Method | Links | |||||||||
|---|---|---|---|---|---|---|---|---|---|---|
| BDE-BiTSEType=DNN-based2025.07 | 3.23 | 3.543 | 3.951 | 25.59 | 3.77 | 4.01 | 1 | 0.126 | 0.484 | |
| Bi-TSE-HRTFVariant=Proposed, Input Clue=HRTF, Type=Complex-valued2025.07 | 3.18 | 3.501 | 3.911 | 21.41 | 3.47 | 3.88 | 0.99 | 0.0756 | 0.0035 | |
| Bi-TSE-HRTF-RIInput Features=RI (Real-valued components), Input Clue=HRTF2025.07 | 3.11 | 3.504 | 3.764 | 21.39 | 2.93 | 3.57 | 0.99 | 0.085 | 0.0044 | |
| BLCMV BeamformerType=Beamforming2025.07 | 3.044 | 3.444 | 3.748 | -0.42 | 3.48 | 3.84 | 0.95 | 0.745 | 0.613 | |
| Bi-TSE-DOAInput Clue=DOA (One-hot encoded), Type=Complex-valued2025.07 | 2.965 | 3.376 | 3.66 | 16.44 | 2.43 | 3.09 | 0.97 | 0.165 | 0.99 | |
| UnprocessedCondition=Raw Input2025.07 | 2.845 | 3.376 | 3.411 | 0 | 1.25 | 1.57 | 0.72 | — | — |