Using AI to Restore Speech in Paralyzed Stroke Survivor After 18 Years

Ann Johnson's life changed dramatically after suffering a brainstem stroke at age 30, which left her paralyzed and unable to speak—a condition known as locked-in syndrome. Despite her paralysis, she retained the ability to think and attempt speech, but her body could not follow her commands. For nearly two decades, Johnson relied on slow, eye-tracking communication tools to express herself. In 2022, as part of a groundbreaking clinical trial led by researchers at UC Berkeley and UC San Francisco, Johnson experienced a remarkable breakthrough using advanced brain-computer interface (BCI) technology. The team developed a neuroprosthesis that reads neural signals associated with speech and translates them into audible words using AI models. This technology enabled Johnson to hear her own voice again for the first time in 18 years, a profoundly emotional moment that highlights the potential of AI-assisted neuroprosthetics. The researchers modeled how speech originates in the brain, identifying the responsible neural pathways, and designed a system that bypasses the physical connection to the body, directly translating thought into speech. The system uses an implant placed over speech-related brain regions, capturing neural activity, which is then processed by AI decoders to produce real-time synthesized speech or facial animations. Improvements continue, with recent innovations reducing the translation delay to about one second, making interactions more natural. Johnson's experience illustrates how such technologies can not only enhance communication but also restore agency and independence for individuals with severe paralysis. Her story inspires future developments toward widely accessible, plug-and-play neuroprostheses, aiming to improve quality of life for many with speech impairments, including stroke survivors and those with neurodegenerative conditions.