Scientists from the University of Cambridge have unveiled a groundbreaking wearable device that uses artificial intelligence to restore natural, fluent speech to individuals who have lost the ability to speak following a stroke.
How the Revoice Collar Works
The innovative device, named 'Revoice', represents a major leap forward in assistive speech technology. Unlike existing methods that can be inefficient or require risky brain implants, this non-invasive collar decodes a wearer's silent speech attempts. It uses sensitive sensors to capture minute throat vibrations and pulse signals as the person tries to talk, even if no sound is produced.
An embedded large language model (LLM) – the same core AI technology that powers chatbots like ChatGPT – then analyses these fragments and cues to predict and reconstruct the intended full sentences. Remarkably, it also works to infer the speaker's emotional state, aiming to deliver speech that sounds like their natural voice.
High Accuracy in Clinical Trials
In an initial trial involving five patients suffering from post-stroke dysarthria, the Revoice system demonstrated impressive accuracy. The research team, led by Professor Luigi Occhipinti from Cambridge's Department of Engineering, reported that the device achieved a sentence error rate of just 2.9 per cent.
Professor Occhipinti explained the profound need for such a solution: "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. That frustration can be profound, not just for the patients, but for their caregivers and families as well."
Future Potential and Development
The researchers believe this technology could extend far beyond stroke recovery. It holds significant promise for aiding individuals with other neurological conditions that affect speech, such as Parkinson's disease and motor neuron disease (MND).
Looking ahead, the Cambridge team aims to develop a multilingual version of the device capable of decoding a wider spectrum of emotional states. However, they caution that extensive clinical trials will be necessary before Revoice can be made widely available to patients.
The full details of this medical AI breakthrough were published in the prestigious journal Nature Communications on Monday, 20 January 2026, in a paper titled 'Wearable intelligent throat enables natural speech in stroke patients with dysarthria'.
