Monday , July 22 2019
Home / australia / Say what! The AI ​​system can decode the brain signals • to the registry

Say what! The AI ​​system can decode the brain signals • to the registry



AI algorithms can process scientists' brain waves and convert them into direct words, according to the new study.

"Our voices help us connect with our friends, our family and the world around us, which is why destroying the power of voice as a result of injury or illness," said Nima Mesgarani, author of the paper published in the scientific report. Columbia University Researcher. "With today's study, we have a potential to recover power, and we will show it, with the right technology, to decipher thoughts of people and to be able to understand them".

Neurons that hear our brain's hearing are excited about the people we hear – or even more people who imagine. How the brain is making fun sound waves or when the facsimile process is built when the people talk about it when it's still unknown to us. However, neurosurences have shown that brain models that are emitted in a task to reconstruct the words that are spoken to each other. It has promoted the idea of ​​building neuroprotectic devices that act as brain-computer interfaces.

Research groups advanced a technique known as a reconstruction of the auditory stimulus through a neural network. First, the audio signal was trained to become a spectrogram, specifying the frequency of sounds in 80 hours of talk time.

The researchers subsequently put electrodes directly into the brain's participation in the epilepsy of electrical activity activity. All of them had normal hearing. They listened to short stories about 30 minutes. The stories were randomly interrupted, and they recited the last scene to prepare their voice reader. The Speaker instructed audible lectures to explain the brain patterns accurately.

The participants listened to a number of 40 numbers – zero and nine – recite. The recorded brain signals were executed by signal loudspeakers, and these samples were reviewed by self-referencing to replicate the numbers to be rebuilt.

Here you can hear an example. It's a bit robotic and impossible, and it's just about nine to nine.

"People can understand and repeat sounds about 75 percent, which is more than the previous ones. Sensitive vibration and neural networks revealed amazing sounds, they told the original sounds, say Mesagarani.

Although intriguing, the experiment is very easy. The system only reconstructs the signals of the listening participants, so they are not their thoughts. It is also a digit recognition, not even integers or phrases. Researchers want to try their own systems with more complicated words, and if they are able to talk or represent people.

"In this scenario, if the user thinks," I need a glass of water, "our system can capture and synthesize brain signals and create verbal words," said Mesagaran. "This would be a game to change. Someone would lose their ability to revive the possibility of connecting to the world via injury or illness." ®


Source link