The device combines ultra-sensitive sensors with artificial intelligence to interpret speech signals and emotional cues, enabling people with speech impairments after a stroke to communicate more naturally.
Worn around the neck, the soft Revoice device records the user’s heart rate and subtle vibrations from throat muscles, then uses these signals to reconstruct intended words and sentences in real time.
The device’s signals are handled by two AI agents: one rebuilds words from fragments of silently mouthed speech, while the other analyzes emotional state and contextual factors, such as time of day or weather, to expand brief phrases into full, expressive sentences.
In a small study involving five patients with dysarthria, a common speech impairment following stroke, the device achieved a word error rate of 4.2 % and a sentence error rate of only 2.9 %.
Unlike existing assistive speech technologies that often rely on slow letter-by-letter input, eye tracking, or brain implants, the Revoice device enables smooth, real-time communication by transforming just a few mouthed words into complete, fluent sentences.
The findings could have a broader impact beyond stroke rehabilitation, potentially benefiting individuals with conditions such as Parkinson’s disease and motor neuron disease. The researchers are now preparing a clinical study in Cambridge involving native English-speaking patients with dysarthria to evaluate the system’s feasibility, which they aim to begin later this year.
Around half of stroke survivors develop dysarthria, or dysarthria combined with aphasia. Dysarthria is a physical disorder that results from weakness in the muscles of the face, mouth, and vocal cords. It affects individuals differently, but often leads to unclear speech, slurred or slowed pronunciation, or speaking in brief, fragmented bursts instead of complete sentences.
When people have dysarthria following a stroke, it can be extremely frustrating for them, because they know exactly what they want to say, but physically struggle to say it, because the signals between their brain and their throat have been scrambled by the stroke. That frustration can be profound, not just for the patients, but for their caregivers and families as well.
Luigi Occhipinti, Professor, Department of Engineering, University of Cambridge
Most stroke patients with dysarthria work with a speech therapist to restore their communication skills, mainly through repetitive speech exercises that involve repeating words or phrases. Recovery timelines differ, but progress typically takes several months and can extend to a year or longer.
Patients can generally perform the repetitive drills after some practice, but they often struggle with open-ended questions and everyday conversation. And as many patients do recover most or all of their speech eventually, there is not a need for invasive brain implants, but there is a strong need for speech solutions that are more intuitive and portable.
Luigi Occhipinti, Professor, Department of Engineering, University of Cambridge
Occhipinti and his team created the Revoice device as a solution of this kind. Sensors within the device detect subtle throat vibrations to capture speech signals and infer emotional states from pulse signals, serving as a simplified yet effective indicator. The device also incorporates an embedded, lightweight large language model (LLM) to predict complete sentences, allowing it to operate using minimal power.
In collaboration with colleagues in China, the researchers conducted a small study involving five stroke patients with dysarthria and ten healthy control participants. During the trial, participants wore the device and silently mouthed short phrases. By nodding twice, they were able to select the option to expand those phrases into full sentences using the embedded LLM.
In one instance, “We go hospital” was expanded into “Even though it’s getting a bit late, I’m still feeling uncomfortable. Can we go to the hospital now?” The sensors in the Revoice device detected that the user was experiencing frustration based on an elevated heart rate and that it was late at night. The LLM used this information to turn three mouthed words into a complete sentence.
Study participants reported a 55 % increase in satisfaction, suggesting that the device could be a promising tool for helping stroke patients regain their ability to communicate.
While large-scale clinical trials are still needed before the device can be widely deployed, the researchers hope future versions will support multiple languages, recognize a wider range of emotional states, and operate as a fully self-contained system for daily use.
This is about giving people their independence back. Communication is fundamental to dignity and recovery.
Luigi Occhipinti, Professor, Department of Engineering, University of Cambridge
Journal Reference:
Tang, C., et al. (2026). Wearable intelligent throat enables natural speech in stroke patients with dysarthria. Nature Communications. DOI: 10.1038/s41467-025-68228-9.