Speech Recognition 벤치마크
Speech Recognition on EasyCom
WER (%) 낮을수록 좋음
- 2022-07-15 — DAJA (MVDR,HMA,1000) (Overlapped Speech): WER (%) 62.36
- 2022-12-21 — ReVISE (bf): WER (%) 52.1
| Rank | Model | WER (%) | Paper | Code | Year |
|---|---|---|---|---|---|
| 1 | ReVISE (bf) | 52.1 | ReVISE: Self-Supervised Speech Resynthesis with Visual Input for Universal and Generalized Speech Enhancement | 2022 | |
| 2 | ReVISE (ch2) | 55.0 | ReVISE: Self-Supervised Speech Resynthesis with Visual Input for Universal and Generalized Speech Enhancement | 2022 | |
| 3 | DAJA (MVDR,HMA,1000) (Overlapped Speech) | 62.36 | Direction-Aware Joint Adaptation of Neural Speech Enhancement and Recognition in Real Multiparty Conversational Environments | 2022 | |
| 4 | Demucs (bf) | 69.8 | ReVISE: Self-Supervised Speech Resynthesis with Visual Input for Universal and Generalized Speech Enhancement | 2022 | |
| 5 | Demucs (ch2) | 86.8 | ReVISE: Self-Supervised Speech Resynthesis with Visual Input for Universal and Generalized Speech Enhancement | 2022 |