Scientists have made a breakthrough in understanding the neural activity behind inner speech, successfully decoding it on command with up to 74% accuracy. The findings, published in the journal Cell, could revolutionize communication for people with severe speech and motor impairments.
Researchers at Stanford University used brain-computer interface (BCI) technologies to decode the neural signals associated with inner speech. They found that when participants thought about speaking, their brain activity showed similar patterns to attempted speech. However, inner speech tended to have a weaker magnitude of activation overall.
The team trained artificial intelligence models to interpret imagined words and demonstrated a proof-of-concept system that could decode imagined sentences from a vocabulary of up to 125,000 words with high accuracy. The BCI was also able to pick up on specific keywords, such as numbers, without prior instruction.
This breakthrough has significant implications for people with disabilities who rely on BCIs for communication. With the ability to decode inner speech, individuals may be able to communicate more easily and naturally. The researchers are exploring ways to refine the system to improve accuracy and expand its capabilities.
The study’s findings were made possible by a collaboration of multiple organizations, including the Assistant Secretary of Defense for Health Affairs, the National Institutes of Health, and the Stanford University faculty.
Source: https://www.eurekalert.org/news-releases/1093888