Metraipolis › archive › dc_d31fa6c7cccf
Myovox: Reading Speech from the Muscles of the Face — arXiv:2609.17548
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2609.17548
- License (as archived)
- CC0-1.0
- Archived
- 2026-10-15T13:49:07.833Z
- SHA-256 of the archived text
e89dec53223722ff99fa8716bacf00110b3f29fa6f43bb80ee158a475885689e- Stable URL of this page
https://metraipolis.com/documento/dc_d31fa6c7cccf
How to cite
Myovox: Reading Speech from the Muscles of the Face — arXiv:2609.17548 — arXiv (en), https://arxiv.org/abs/2609.17548 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_d31fa6c7cccf (Metraipolis archive dc_d31fa6c7cccf, sha256 e89dec53223722ff99fa8716bacf00110b3f29fa6f43bb80ee158a475885689e)
BibTeX
@misc{dcd31fa6c7cccf,
title = {Myovox: Reading Speech from the Muscles of the Face — arXiv:2609.17548},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2609.17548},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_d31fa6c7cccf; sha256 e89dec53223722ff99fa8716bacf00110b3f29fa6f43bb80ee158a475885689e},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.17548. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
Myovox, from myo (muscle) and vox (voice), decodes open-vocabulary English text from 31-channel surface electromyography (sEMG) recorded from the muscles of the face during vocalized speech. It takes the single-subject emg2speech General Corpus from a published 51.17% word error rate to 18.53%, in three separable moves, each measured in isolation. First, I recover the open-vocabulary decode settings missing from the public release and reach a faithful 40.63% WER / 39.02% PER baseline whose phone error rate matches the published one to within 0.8 points, so the acoustic model is reproduced faithfully. Second, I replace the causal encoder with a bidirectional Conformer trained by a four-term cross-modal distillation against the parallel audio's WavLM-Large layer-9 features, reaching 26.14% WER / 22.34% PER from the electromyography alone. Third, I ensemble two acoustic models, union their multi-scale n-best lists, and rerank with a QLoRA-fine-tuned 7B language model, reaching 18.53% WER, the best result reported on this corpus, though not the best reported for sEMG-to-text on other corpora (Section 2). I then report the negative result that bounds the whole approach: reranking is exhausted at 18.5% because the binding constraint is the electromyographic acoustic phone error rate (~20.9%), not the language model. The correct words are simply absent from the acoustic posteriors, so no reranker can reach the 9.30% n-best oracle. All test numbers are on the 400-sentence held-out test set under the authors' official 8,500 / 760 / 400 sequential split; every hyperparameter is tuned once on validation and applied once to test. Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-09-17 in cs.CL). Categorie arXiv: Computation and Language (cs.CL). Pagina del lavoro: https://arxiv.org/abs/2609.17548 — 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 2026-10-15T13:48:24.257Z dalla pagina https://arxiv.org/catchup/cs.CL/2026-09-17?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).