Target Speaker Extraction on ICASSP DNS-challenge Track 2 - Speakerphone 2023 (test)
4.11SIG ScoreLLaSE-G1
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| LLaSE-G1Type=Generative, Inference Scaling=multi2025.03 | 4.11 | 3.86 | 3.58 | |
| LLaSE-G1Type=Generative, Inference Scaling=single2025.03 | 4.08 | 3.84 | 3.55 | |
| NoisyType=Unprocessed Baseline2025.03 | 4.05 | 2.16 | 2.5 | |
| TEA-PSE 3.0Type=Discriminative2025.03 | 3.99 | 3.95 | 3.49 | |
| NAPSEType=Discriminative2025.03 | 3.92 | 4.17 | 3.56 |