How artificial intelligence gave a paralyzed woman her voice

Thanks to a revolutionary brain implant and a digital avatar, a speechless stroke victim can express herself again.

At age 30, Ann, a math teacher in Canada, suffered a stroke that left her paralyzed and unable to speak. Eighteen years old, a victim of locked-in syndrome, she laboriously communicated.

The computer itself allows him to produce a metallic voice at the rate of fourteen words per minute. Until his participation in research conducted by the Bravo3 team, led by Edward Chang, chair of neurological surgery at the University of California, San Francisco, as part of a brain-computer interface procedure.

The system can decipher 80 words per minute

Electrodes implanted in Anna’s brain and connected to a set of computers via a cable attached to her skull pick up the electrical signals generated by her brain when she tries to move her face.

Although his muscles are paralyzed, the electrical signals are still present, but they cannot reach the muscles to activate them. This is where artificial intelligence (AI) comes into play. The signals captured by the implant are transmitted to a computer equipped with a deep learning algorithm. The AI ​​learns to read Anna’s intentions, even though her body can’t express them.

The process requires weeks of training. Ann repeats sentences with a vocabulary of 1,024 words until the computer recognizes the brain patterns associated with all the basic sounds of her speech. Once the AI ​​interprets the signals, it translates them into facial movements for a digital avatar, capable of representing Anna’s facial expressions and her voice, the timbre of which was recreated by modeling her vocal frequencies that were recorded before her stroke. in 2005. The goal is to eventually develop a wireless device. But for her 18-year-old daughter, who is hearing her mother’s voice for the first time, the process is already unbelievable.

Leave a Comment