Losing your voice in a matter of minutes, while keeping your thoughts perfectly clear: this is the reality for many stroke survivors. The brain wants to speak, the mouth can no longer do so, loved ones guess without always understanding. Faced with this scene, engineers from the University of Cambridge have imagined a surprising device, which is starting to change the situation.
Giving voice to patients after a stroke
About half of the people affected by
Stroke develop a
dysarthriathis speech disorder where the muscles of the face, mouth and vocal cords are weakened. Its manifestations vary from person to person, but it often results in difficulty speaking clearly, slow or slurred speech, or producing words in fragments, rather than complete sentences. Even with speech therapy, speaking spontaneously remains difficult.
“After some practice, patients are usually able to perform the repetitive exercises, but they often have difficulty with open-ended questions and everyday conversations.”explains Professor Luigi Occhipinti from the Department of Engineering at the University of Cambridge who led the study. “Because many patients eventually regain most, if not all, of their speech, invasive brain implants are no longer necessary. On the other hand, there is a significant need for more intuitive and portable voice solutions.”.
And Professor Occhipinti’s team today offers a simple
choker which could give these patients a voice again.
Revoice, an AI-boosted collar for speaking
The device Returndescribed in the journal
Nature Communicationsappears as a soft and flexible necklace worn around the neck. Ultra-sensitive sensors integrated into the textile record both the tiny vibrations of the throat muscles when the patient articulates silently, and their heart rate. These signals are then analyzed byartificial intelligence to reconstruct, in real time, words and then entire sentences. All without surgery.
This assistance targets an experience that caregivers know well. “When a person suffers from dysarthria after a stroke, it can be extremely frustrating because they know precisely what they want to say, but struggle physically to express it because the signals between their brain and throat are disrupted by the stroke.“, underlines Professor Luigi Occhipinti from the Department of Engineering at the University of Cambridge who led the study. ” This frustration can be profound, not only for patients, but also for their caregivers and families.“.
Throat and heartbeat vibrations help guess sentences
Concretely, the person suffering from a stroke wears the necklace and forms words silently. The sensors record very short sequences of vibrations on the throat and send them to a first AI module which recomposes the words then the basic sentence. A second module adds the emotion deduced from the pulse and some context information (such as the time of day or weather conditions), which makes it possible to obtain a complete, natural and personalized sentence.
© Nature
During the trials, five patients with dysarthria tested
Return. By simply articulating “We’re going to the hospital“, one of them saw the system produce: “Even though it’s late, I don’t feel well. Can we go to the hospital now?“. The Revoice device’s sensors deduced that the user was frustrated by their high heart rate and that it was late. The device was able to use this data to turn three spoken words into a complete sentence.
The device showed 4.2% errors on words and 2.9% on sentences, participant satisfaction increased by 55% and a voice response in one second.
Unlike current voice assistant technologies, which often require laborious letter-by-letter typing, use of eye tracking or brain implants, the Revoice device enables seamless, instantaneous communication, converting a few articulated words into complete, natural-sounding sentences.
Objective: restore autonomy to many patients
The researchers point out that Return remains a prototype and that larger clinical trials will be required before large-scale distribution. A study is planned in Cambridge with English-speaking patients, with the aim of then adding multilingual capabilities and sufficient autonomy for continuous use over an entire day. The team also imagines helping people suffering from Parkinson’s disease or Charcot’s disease by restoring more autonomy and dignity to communication.
Although detailed clinical trials are essential before widespread commercialization, researchers envision that future versions of the device will include multilingual capabilities, broader recognition of emotional states and full autonomy for everyday use.