SepRQ extends the BestRQ self-supervised framework with a pseudo-source-separation objective.
Mask-free separation pretraining improves representations of overlapping speech
SepRQ learns speaker-attributed representations at multiple time scales and reportedly outperforms established speech models across separation, diarization and target-speaker benchmarks.
Independent
Séverin Baroudi · Hervé Bredin · Ricard Marxer
Research Digest··2 min read
Baroudi, Bredin and Marxer introduce SepRQ, a self-supervised speech model designed specifically for mixtures containing multiple speakers.
Why this paper
Independent
In one line
SepRQ achieves state-of-the-art multi-speaker representation learning using mask-free pseudo-source-separation with frozen codebooks.
What we could check
- ·No code link found
- ·No weights link found
- ·No dataset link found
- ·No compute details found
- ✓Limitations stated by the authors
- ·No benchmark numbers found
Observed from the paper text and links we have. Absence here means we did not find it, not that it does not exist.
§