Thanks to advances in artificial intelligence (AI), American researchers have given voice to a woman, paralyzed because of Charcot’s disease.
Speech is a gift many of us take for granted. But for those deprived of it due to paralysis, a study published in the journal Nature August 23 offers new hope. Thanks to artificial intelligence (AI) and neural implants, people with facial paralysis may soon regain the ability to speak with others. American researchers have developed brain-computer interface (BCI) technologies that allow paralyzed people to communicate faster and more accurately. In other words, these brain implants decode their thoughts to make it into words.
“Three out of four words are deciphered correctly”
A team from Stanford University collaborated with Pat Bennett, a 67-year-old woman with Charcot’s disease (or amyotrophic lateral sclerosis) and who had lost ability to speak due to illness. His brain, however, continued to send signals indicating his intention to speak. They implanted mini-electrodes in areas of the brain involved in speech and used algorithms to interpret neural activity as she attempted to speak sentences. L’AI then decoded the words from the phonemes. “This system is designed to know which words must precede the others and which phonemes make up which words“, said Dr. Frank Willett, neuroscientist at Stanford University and co-author of the study. Result? The interface decoded speech at a speed of 62 words per minute with a word error rate of 9.1% for a vocabulary of 50 different words. “About three out of four words are deciphered correctly”, commented Dr. Willett during the press conference. For him “It is possible to imagine a future in which we will be able to restore a fluent conversation to a paralyzed person“.
“For those who don’t speak, it means they can stay connected to the wider world, maybe keep working, nurture relationships friendly and family” underlined Pat Bennett. In another study, the University of California at San Francisco worked with Ann, a 47-year-old woman, mute since a stroke. Rather than implanting electrodes, they used a less invasive technique called “electrocorticography (ECoG)”. Their BCI produced 78 words per minute with an error rate of 25.5% for a vocabulary of 1024 words. The researchers also created algorithms to convert Ann’s brain signals into a synthetic voice, and they customized the voice to sound like Ann’s. “Just hearing a voice like yours is moving” said Ann. Although these breakthroughs are revolutionary, many improvements are still needed. The devices also need to be tested on more people to prove their reliability. An ideal BCI would be completely implantable and wireless, avoiding any complications due to cables or connectors.