Particle.news
Download on the App Store
5 ARTICLES
·
last mo.

Breakthrough Brain-Computer Interface Enables Near Real-Time Speech for Paralyzed Patients

UC Berkeley and UCSF researchers unveil a system that decodes neural signals into naturalistic speech, restoring communication for individuals with severe paralysis.

Image
Image
Image
machine learning artificial intelligence AI speech generative ai neuralink brain-computer interface UC Berkeley UC San Francisco

Overview

  • The new brain-computer interface (BCI) system reduces speech synthesis latency to under one second, enabling fluid, near real-time communication.
  • Tested on a stroke patient unable to speak for 18 years, the system successfully synthesized speech using her pre-injury voice, enhancing her sense of embodiment.
  • The technology uses neural signals from the motor cortex and a deep learning model to decode brain activity into audible speech with high accuracy.
  • Researchers demonstrated the system's adaptability, including its ability to generalize beyond training data and synthesize rare words like those in the NATO phonetic alphabet.
  • Future research aims to enhance expressivity by incorporating tone, pitch, and loudness, while exploring broader applications, including non-invasive brain-sensing technologies.