Can’t speak clearly after a stroke? A wearable turns tiny throat and body signals into full sentences, helping people talk again easily.
Many stroke survivors cannot speak clearly. Their thoughts remain intact, but weak throat and face muscles break speech into slow fragments. Existing aids often force them to type, use eye tracking, or rely on implants. These methods are slow, tiring, and disrupt conversation.
To address this, researchers at the University of Cambridge in the UK developed Revoice, a wearable device worn around the neck that helps users speak again in real time. It is designed for people with dysarthria after stroke, a condition that affects about half of all stroke survivors.
Revoice focuses on how these users already try to communicate—by forming short, broken phrases. Instead of asking them to spell words, the device listens to throat muscle signals and body cues to rebuild what the person wants to say.
The device senses vibrations from throat muscles and tracks heart rate. These signals capture speech intent and emotional state. Using this input, the system reconstructs words as the user silently mouths them.
Two AI systems work together. One converts silently mouthed words into text. The other uses context—such as emotion, time of day, or situation—to expand short phrases into full sentences. For example, the phrase “We go hospital” can be turned into a complete sentence that fits the moment.
In a study with five stroke patients with dysarthria, the system achieved a word error rate of 4.2% and a sentence error rate of 2.9%. Participants reported a 55% increase in satisfaction while communicating with the device.
The study also included ten healthy participants to validate the approach. Users could choose to expand phrases by nodding twice. All processing ran on an embedded language model inside the wearable.
Unlike current speech aids, Revoice does not depend on keyboards, eye tracking, or brain implants. It allows conversation by predicting full sentences from minimal input. The researchers believe the same approach could help people with Parkinson’s disease and motor neuron disease, where speech muscles weaken over time.
