US researchers have developed a brain implant that can convert a paralysed woman’s thoughts into speech almost instantly, marking a breakthrough in brain-computer interface (BCI) technology.
The experimental device, which uses artificial intelligence, builds on previous efforts to restore communication for individuals who have lost their ability to speak. A team in California had previously managed to translate the thoughts of Ann, a 47-year-old quadriplegic woman, into speech, but with an eight-second delay. The latest advancement reduces that delay to near real-time, allowing her thoughts to be vocalised in just 80 milliseconds.
“Our new streaming approach converts her brain signals to her customised voice in real time, within a second of her intent to speak,” said Gopala Anumanchipalli, senior author of the study published in Nature Neuroscience.
Ann, a former high school math teacher who has been unable to speak since a stroke 18 years ago, was shown sentences on a screen, which she then repeated in her mind. The system then recreated her voice using deep-learning AI trained on her past speech recordings.
The breakthrough has brought Ann one step closer to her dream of becoming a university counsellor. Ann was “very excited” to hear her old voice again, according to the researchers. “She reported a sense of embodiment,” Anumanchipalli said.
The AI-powered system interprets brain signals after a person has decided what to say but before their vocal muscles attempt to move, explained co-author Cheol Jun Cho. However, while promising, the technology still has limitations. The model has a vocabulary of only 1,024 words and is not always fully accurate.
Patrick Degenaar, a neuroprosthetics professor at Newcastle University who was not involved in the research, described the study as “very early proof of principle” but still “very cool.”
Unlike Elon Musk’s Neuralink, which uses electrodes that penetrate the brain, this system relies on an external array of electrodes, making it easier to deploy in hospitals. With adequate funding, Anumanchipalli estimates that this technology could help people communicate within five to ten years.
