Brain Implant Lets Paralyzed Patients Speak and Gesture via Avatar — SkimNews
Get the Health newsletter
Daily health & science — research, biotech, public health, the studies worth knowing. Free.
- Samantha Brosler and colleagues at the University of California, San Francisco tested an iPhone-sized brain implant in two paralyzed participants — Bravo-1r (stroke survivor) and Bravo-6 (ALS patient) — both with near-total loss of limb movement and intelligible speech.
- The implant spans a large portion of the sensorimotor cortex, enabling simultaneous decoding of attempted speech and gestures — a combination prior brain implants had not successfully achieved.
- Custom machine learning models were trained on brain recordings of participants attempting 10 phrases (e.g., "hello," "how's it going?") and 10 gestures (waving, clapping, head-shaking), then tested on novel pairs displayed on screen.
- Bravo-1r's gestures and speech were decoded at 88% and 84% accuracy; Bravo-6's at 66% and 70%, compared with a roughly 9% chance baseline.
- Henri Lorach at the University of Lausanne called the results "a very nice piece of work" with robust findings but said the limited vocabulary and lower accuracy must improve for "daily use without frustration."
- Future iterations could move a person's own body and produce a synthetic voice rather than displaying decoded output through an avatar, Brosler said.
Why it matters: The proof-of-concept decodes gestures and speech at 66–88% accuracy, but the system recognizes only 10 phrases and 10 gestures, and an outside reviewer flagged that daily-use deployment requires substantially higher accuracy and broader vocabulary — meaning paralyzed users gain a meaningful validation of simultaneous decoding, not yet a practical communication tool.
Ask SkimNews
