Researchers from the University of California, Berkeley and San Francisco have developed a new brain-computer interface that restores “naturalistic speech for people with severe paralysis.” In a study published by Nature Neuroscience, the innovation addresses a huge challenge in speech neuroprostheses and is a huge step forward since it allows people who lost the ability to speak to engage in real-time communication.
Researchers say they used artificial intelligence to develop a streaming method that translates brain signals into audible speech almost instantly. The method is similar to how voice assistants like Alexa and Siri quickly decode speech, resulting in “near-synchronous voice streaming” that sounds natural.
According to the study’s co-load author Cheol Jun Cho, neuroprotheses works by sampling neural data from the motor cortex, a part of the brain that controls speech and then uses AI to decode these brain signals into speech. Researchers say they tried this on a 47-year old woman named Ann, who has been unable to speak for the last 18 years.
In the clinical trial, Ann was implanted with electrodes on her brain surface, which recorded neural activity when she tried to speak sentences on a screen. The system then decoded these signals into an audible speech, which was powered by an AI model that replicated her pre-injury voice.
While brain computer interfaces have been common for quite some time, one of the major breakthroughs in the study was the near real-time translation of thoughts into audio. Previously developed BCI models often suffered from a long latency, which, on average, resulted in an 8-second delay for a single sentence. The newly developed technology can help improve the lives of people who are affected by conditions such as ALS or stroke-induced paralysis.
© IE Online Media Services Pvt Ltd
