Metraipolis › archive › dc_4fc754d5315c

Test-time adaptation for speech enhancement with an autoregressive speech prior — arXiv:2609.03622

Source
arXiv (en)
Original
https://arxiv.org/abs/2609.03622
License (as archived)
CC0-1.0
Archived
2026-10-17T01:12:00.858Z
SHA-256 of the archived text
de99c5f880367f16272e54066d818f216c730a7d90b018144fdba45b1febcc5c
Stable URL of this page
https://metraipolis.com/documento/dc_4fc754d5315c

How to cite

Test-time adaptation for speech enhancement with an autoregressive speech prior — arXiv:2609.03622 — arXiv (en), https://arxiv.org/abs/2609.03622 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_4fc754d5315c (Metraipolis archive dc_4fc754d5315c, sha256 de99c5f880367f16272e54066d818f216c730a7d90b018144fdba45b1febcc5c)
BibTeX
@misc{dc4fc754d5315c,
  title = {Test-time adaptation for speech enhancement with an autoregressive speech prior — arXiv:2609.03622},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2609.03622},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_4fc754d5315c; sha256 de99c5f880367f16272e54066d818f216c730a7d90b018144fdba45b1febcc5c},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.03622. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Test-time adaptation (TTA) offers a promising direction for improving speech enhancement models under mismatched acoustic conditions, without requiring access to labeled target data. In this work, we propose a single-utterance TTA method that regularizes a pretrained speech enhancement model using an autoregressive prior trained on clean speech latent representations extracted from a neural audio codec. Adaptation is performed by minimizing the Kullback-Leibler divergence between the enhanced speech distribution and the clean speech prior. Experiments across multiple noisy speech datasets show consistent improvements in speech quality, particularly under training-testing noise mismatch conditions. Code and audio examples are available online.

Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-09-04 in cs.AI).
Categorie arXiv: Sound (cs.SD) ; Artificial Intelligence (cs.AI).
Pagina del lavoro: https://arxiv.org/abs/2609.03622 — autori e testo completo sono lì (qui omessi i nomi: nessun dato personale nel deposito).
Licenza: titolo e abstract sono metadati arXiv distribuiti in CC0 1.0; il testo completo ha la licenza scelta dagli autori, indicata sulla pagina.
Letto il 2026-10-17T00:28:02.988Z dalla pagina https://arxiv.org/catchup/cs.AI/2026-09-04?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).