[ad_1]
Scientists at Columbia University in the United States have created a technology that can badyze brain activity and transform it into clear, intelligible language – an important scientific step in creating a system capable of read the thoughts.
The research, published Tuesday in the journal Nature, helped treat the words heard by patients and reconstruct them with an artificial voice using speech synthesizers and artificial intelligence.
According to the researchers, this technique could create new ways for computers to communicate directly with the brain – and to reproduce in the future not only what patients have heard, but also what they have heard.
Photo: DW / Deutsche Welle
they think.
The goal of technology is to help people who have lost their speech find the ability to communicate with the outside world. This includes victims of stroke or suffering from diseases such as amyotrophic lateral sclerosis (ALS), which was endured by British physicist Stephen Hawking.
"The voice allows us to communicate with friends, family and the world around us.Losing the floor is so devastating," says neurologist Nima Mesgarani, one of the authors of the research. "With this study, there is a possible way to restore this power.We have shown that with appropriate technology, the thinking of these people can be decoded and understood by any listener."
To reproduce the speech, scientists used an instrument called vocoder, a computer algorithm capable of synthesizing the voice after being trained to record speaking people. "It's the same technology used by Amazon Echo or Apple's Siri to give verbal answers to our questions," Mesgarani said.
To teach the vocoder to interpret brain activities, the Columbia University team experimented with patients with epilepsy and undergoing brain surgeries – so their skull was open.
"We asked these patients to hear sentences pronounced by different people while measuring patterns of brain activity," says Mesgarani. "These neuron models formed the vocoder."
Next, these same patients heard a recording counting from 0 to 9, while their cerebral signals were recorded and sent to the vocoder via neural networks – a type of artificial intelligence capable of reproduce the neural structure of a human brain. .
The result was a robotic voice reciting the same numbers heard by the patients. "The sensitive vocoder and the powerful neural networks reproduced the sounds originally heard with surprising accuracy," says Mesgarani. On the Columbia University website, you can hear what looks like a robot with 0-9 in English.
The neuroengineering team now wants to try to reproduce more elaborate words and sentences, and then do the same tests using the cerebral signals emitted when a person is talking or thinking about doing so.
According to Mesgarani, the ultimate goal is to turn this technology into a system that can be implanted in the patient's brain and translate thoughts into words.
"In this scenario, if the patient thinks" I need a glbad of water ", our system will be able to capture the cerebral signals generated by this thought and transform them into synthesized verbal speech. " "This would be a watershed, which would give anyone who has lost the ability to speak, whether as a result of an injury or illness, a new chance to connect to the world that has them." 39; surrounds ".
