[ad_1]
The site says in a report translated by "Arab"The researchers adapted computers to read words directly into people's brains by teaching artificial neural networks how to convert neuron recordings into sounds.
"Three different teams of researchers have been able to identify words through neural recordings," said Kelly Curwick, scientific columnist for Science magazine.
In each study, the electrodes placed directly on the brain recorded the neuronal activity of patients undergoing brain surgery as they listened to or read certain words, and the researchers then tried to identify them. Thus, the researchers were able each time to transform the electrical activity of the brain into partially understood voice data.
The first study, published in Biorexive on October 10, 2018, included an experiment in which researchers made audio recordings of epileptic patients undergoing brain surgery. (The neuronal recordings documented in this experiment must be very detailed for their badysis to be possible, but such details are rarely available when the brain is exposed during surgery so that researchers can place electrodes directly on them.)
When hearing patients, researchers recorded the activity of neurons in brain regions responsible for sound processing. In fact, scientists have tried many different ways to turn this neural activity into clear speech and have discovered that artificial intelligence is the best solution. For example, when the researchers entered the results data for eleven listeners in the "Voice Encoder" program, a device that generates human voices, these people were able to correctly identify the words at 75%.
The second study, published November 27, 2018, was based on the nerve records of some people undergoing surgery to eliminate brain tumor. As patients read unambiguous words aloud, the researchers recorded participant and neuronal activity in the brain regions responsible for speech production. As a result, these researchers adapted artificial neural networks to convert neuron recordings into real sounds, the results being somewhat similar to those produced by microphones.
The third study, published on August 9, 2018, recorded the activity of the brain region responsible for converting words into muscle movements. In this context, the researchers found that they were able to extract whole sentences from the brain (also recorded during an operation of the brain in patients with epilepsy), and then these Recordings were shown to some people for interpretation, who managed to know the words at 83%. This experiment identified patterns badociated with the production of individual sections rather than whole words.
The purpose of all these experiments is to allow people who have lost the ability to speak via a computer-brain connection, but science has not reached this stage yet, the website said. The journal "Science" has reported that interpreting a person's speech patterns by imagining is more complex than interpreting neural patterns by listening or speaking.
In conclusion, the website indicates that it is important to emphasize that these studies are limited, the first relying on data from only five patients, the second examining the results of six patients, while the third only covered three patients and all these records did not last more than one hour. . However, science continues to progress and artificial speech devices that communicate directly with the brain may one day become real.
To read the original text, click here.Right here)
[ad_2]
Source link
