'Revoice': Cambridge's AI choker helps stroke survivors speak again

Researchers at the University of Cambridge have developed Revoice, an AI-powered wearable collar designed to help stroke survivors with dysarthria regain natural speech without invasive brain implants. The device uses ultra-sensitive sensors to capture throat vibrations and heart rate, combined with AI algorithms that reconstruct silently mouthed words and interpret emotional and contextual cues to form complete, fluent sentences. In tests involving five stroke patients, Revoice achieved a low word error rate of 4.2% and a sentence error rate of 2.9%, with participants reporting a 55% increase in user satisfaction. Revoice offers a practical, lightweight, and washable solution for everyday use, running on an energy-efficient embedded language model. Unlike traditional methods that rely on slow, letter-by-letter communication or invasive implants, this device facilitates real-time, natural-sounding speech recovery. Beyond stroke rehabilitation, the technology holds potential for aiding individuals with other neurological conditions such as Parkinson’s disease and motor neuron disease. The research, led by Professor Luigi Occhip