Scientists Develop Al That Turns Speech to Text

The University of California’s scientists has recently been working on a system, which could ultimately be used to convert brain activity into text in patients who cannot speak or type.

The brain-reading machine will use brain implants to read what neurons do when someone speaks. An algorithm then uses the information to create a string of numbers, which another system converts into simple text.

“We are not there yet but we think this could be the basis of a speech prosthesis,” said co-author on the paper Dr Joseph Makin from UCSF. The researchers (Makin and his colleagues David A. Moses and Edward F. Chang) tested the system on four people with brain implants that monitor their epileptic seizures.

The participants had to repeatedly read out loud 50 different sentences while their brain signals were fed into a computer and decoded into text. The Al was significantly more accurate than other speech transcribing machines, although it still struggled to translate some phrases.

The researchers stressed that the system currently only works when someone speaks aloud, not much help for people who can’t speak at all. However, it could be improved to benefit people who can’t communicate verbally, such as those suffering from locked-in syndrome, a brain-related disorder that causes paralysis.

Sources: paper, theguardian.com, arstechnica.com

Comments are closed.