Overview
- The soft, washable neck choker captures tiny throat muscle vibrations and heart‑rate signals to reconstruct words and generate fluent sentences in real time.
- Two AI components handle silent speech decoding and interpret emotional and contextual cues, with sentence expansion triggered by a double‑nod gesture.
- In tests with five stroke patients with dysarthria, Revoice recorded a 4.2% word error rate and a 2.9% sentence error rate, and users reported a 55% satisfaction increase.
- The technology remains at an early stage and is not clinically available, with a larger trial for English‑speaking patients planned in Cambridge this year and goals for multilingual support and more robust emotion detection.
- The University of Cambridge led the work with collaborators in China, funding from the British Council, Haleon, and EPSRC/UKRI, and cautious support from the Stroke Association noting the preliminary evidence.