9621 Agnes Crossing, Lake Suzanneview, New Mexico Island 84604-9295.
Scientists in California, USA, have created a brain-computer interface (BCI) that can decode speech at up to 62 words per minute. The study aims to help people with
Scientists in California, USA, have created a brain-computer interface (BCI) that can decode speech at up to 62 words per minute.
The study aims to help people with speech problems, but it still needs to be peer-reviewed.
Published January 21 on the preprint server bioRxivthe article describes this study, which aims to analyze brain activity in a small area of the cerebral cortex to convert them into connected speech using a machine learning algorithm.
Perovskite solar cell makes reactor 100 times more powerful than other models
By achieving their goal, researchers will be able to give a voice to people who can no longer speak, especially those who have suffered from diseases such as paralysis and amyotrophic lateral sclerosis (ALS). There are already some typing-based solutions to help these people, but a speech interface that can work in conjunction with the brain can speed up this type of decoding.
“Here we demonstrate speech BCI that can decode unlimited sentences from a large vocabulary at 62 words per minute, for the first time that BCI far exceeds the communication speed that alternative technologies can provide for people with paralysis, such as ocular palsy. tracking“, the researchers said.
With this recurrent neural network decoder, researchers hope to convert signals emitted by the brain into words very quickly. In one experiment with an ALS patient who could move his mouth but had difficulty speaking, they were able to register activity in two small areas of the brain. This excited the researchers because the combination of orofacial movements associated with neural activity “is probably strong enough to maintain speech BCI despite paralysis and narrow coverage of the cortical surface.”
Clearly there is still much to be improved as part of the ongoing study and the Recurrent Neural Network (RNN) decoder error rate was around 20%.
“Our demonstration is evidence that decoding pre-speech movements from intracortical recordings is a promising approach, but it is still not a complete and clinically viable system.‘, the researchers concluded in the article.
As stated above in the text, the study has not yet been peer-reviewed, and therefore the researchers hope to optimize the algorithm used in the study by examining other areas of the brain.
Mundo Conectado Deal Center: selection of discounts and lowest prices
Best deals on electronics, cell phones, TVs, soundbars, drones and more
Source: bioRxiv
…..
Source: Mundo Conectado
I’m Maurice Knox, a professional news writer with a focus on science. I work for Div Bracket. My articles cover everything from the latest scientific breakthroughs to advances in technology and medicine. I have a passion for understanding the world around us and helping people stay informed about important developments in science and beyond.
