A woman paralyzed by a stroke speaks again through a digital avatar that reproduces her voice and even her facial expressions: credit goes to a new brain-computer interface that, for the first time, manages to translate brain activity signals into words and phrases to generate faster and faster communication to enable more natural communication. The results of the study will be published on natures by researchers at the University of California at San Francisco and Berkeley.

A film with 253 electrodes

The team led by the neurosurgeon Edward Chang, attached a very thin rectangular film with 253 electrodes to the woman’s brain: placed over an area crucial for communication, the device detects the brain signals that can no longer be sent to the phonatorium, the mouth, due to the stroke. larynx) and facial muscles. The signals are then sent to an artificial intelligence system that has been trained for weeks to recognize the woman’s brainwaves that correspond to the different sounds (phonemes) that make up words.

A synthesizer with the patient’s voice

To make everything more realistic, the researchers combined an algorithm that synthesizes words with the patient’s own voice (taken from some recordings made at her wedding). Finally, the avatar’s face reproduced on the computer screen was animated based on the woman’s brain signals, thanks to software that reproduces the movement of facial muscles.

During a stroke, two million neurons die every minute because time is precious