Speech Reconstruction on Full-band SR 16kHz
90.26STOIMSRHuBERT
Evaluation results
| Method | Links | |
|---|---|---|
| MSRHuBERTPre-training sampling rate=Multi-rate2026.03 | 90.26 | |
| Re. 48kHz HuBERT BasePre-training sampling rate=48kHz2026.03 | 89.75 | |
| Re. 24kHz HuBERT BasePre-training sampling rate=24kHz2026.03 | 89.61 | |
| Re. 16kHz HuBERT BasePre-training sampling rate=16kHz2026.03 | 89.35 | |
| Re. 16kHz HuBERT Base (w/o resampling)Resampling in fine-tuning=No2026.03 | 89.35 | |
| HuBERT BasePre-training sampling rate=16kHz2026.03 | 88.46 |