Brain chip that translates thoughts into speech
Technology with the ability to decode a person’s inner thoughts and translate them into speech in real time has moved a step closer after the testing of a new brain implant, scientists have claimed.
Researchers at Stanford University in California said that their device was able, with 74 per cent accuracy, to detect a person’s “inner speech” — where they imagined saying a specified set of words in their mind without any effort to physically utter them — and decipher which words they were thinking about.
They said that it could represent a breakthrough for people with various forms of paralysis or impairments that prevent them from speaking unaided and could one day “restore communication that is as fluent, natural and comfortable as conversational speech”.
The researchers said that it could be a more effective system than existing forms of brain implant, which ask patients to attempt to form sounds and then analyse the movement of their vocal and facial muscles to decode what they are trying to say.
It would also be much faster than older technologies, such as those that track a user’s eye movements to select letters and type out words, they said.
“If you just have to think about speech instead of actually trying to speak, it’s potentially easier and faster for people,” Benyamin Meschede- Krasa, a PhD student and an author of the study, published in the journal Cell, said.
The study involved four people with severe paralysis from either amyotrophic lateral sclerosis (ALS) or a brainstem stroke and asked them to either attempt to speak or to imagine speaking a set of words.
They implanted tiny electrodes in the motor cortex of their brains, which controls speaking. They found that similar brain areas were activated when both imagining speech and actually attempting to speak, though the imagined speech produced weaker signals.
In a small proof-of-concept study, the device was able to identify pre-specified sentences created from a 125,000-word vocabulary with an accuracy rate of up to 74 per cent, powered by an artificial intelligence model.
The implant, part of a brain-computer interface (BCI), was also able to detect some words that the participants had not been instructed to say, including numbers when the participants were counting shapes on a screen.
This could prompt fears of systems being able to read a person’s thoughts against their wishes, so researchers also tested whether the users could activate the system by imagining a password and found that it worked with 98 per cent accuracy.
It is still not able to translate freeflowing speech without making significant errors, but researchers hope that improvements in the devices and the AI system will one day make that possible.
“The future of brain-computer interfaces is bright,” said Frank Willett, the study’s senior author. “This work gives real hope that speech BCIs can one day restore communication that is as fluent, natural, and comfortable as conversational speech.”
No comments:
Post a Comment