AI wearable gives a voice to people rendered speechless by strokes
January 21, 2026
The Revoice device uses two AI agents to detect the wearer’s silently mouthed speech and emotional state, outputting speech via a synthetic voice module
University of Cambridge
View 4 ImagesView gallery – 4 images
Speech impairment, known as dysarthria, is a very common consequence of strokes, affecting nearly half of all survivors. A stroke can cause weakness in the facial muscles and vocal cords, making it difficult to speak fluently, clearly, or in full sentences. While most people eventually recover, the process is often slow and frustrating, and it significantly affects quality of life during rehabilitation.
Scientists at the University of Cambridge have been working to change this. Building on previous research, they have developed a wearable device called Revoice, designed to help people with dysarthria communicate more naturally.
“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,” explains Prof. Luigi Occhipinti, one of the leaders of the research team.
Reading and reconstructing these signals is exactly what Revoice is intended to do. Since many patients eventually regain natural speech and only need temporary assistance, invasive solutions such as brain implants are often unnecessary. Revoice offers a non-invasive, wearable alternative.
The device looks like a soft, adjustable choker with textile strain sensors and a wireless printed circuit board. Its integrated AI system decodes speech signals using two AI agents: one reconstructs words from silently mouthed speech and predicts phrases by reading vibrations in the throat muscles, while the other detects the emotional state of the patient by measuring neck pulse. This allows the device not only to reconstruct full sentences but also to make them emotionally resonant and logical.
Earlier silent-speech systems were mostly tested on healthy participants and lacked real-world application. They also required users to pause for one to three seconds between words, creating awkward unnatural delays in speech. Revoice overcomes those limitations by using an AI-driven throat sensor system and a lightweight large language model to convert mouthed words into complete sentences instantly.
After initial trial with healthy participants, the device went through some optimization and was later tested on five stroke patients, showing impressively low error rates of 4.2% for words and 2.9% for sentences. In one example, a patient mouthed the phrase “We go hospital,” which Revoice converted into “Even though it’s getting late, I’m still feeling uncomfortable. Can we go to the hospital now?”. Participants reported a 55% increase in user satisfaction and confirmed that the device enabled them to communicate as fluently as they did before the stroke.
The researchers believe that beyond stroke patients, the device could also help people with Parkinson’s disease and motor neuron disease.
Revoice is made from durable, breathable, and washable fabric, making it practical for daily use. It is powered by an 1,800-mWh battery, which is expected to last an entire day on a single charge. Before the product can reach market, it will need to undergo more extensive clinical trials. If successful, the research team plans to add support for multiple languages and a wider range of emotional expressions.
The study is described in a paper that was recently published in the journal Nature Communications.
Source: University of Cambridge
View gallery – 4 images
We recommend
canisgallicus 