Particle.news

Download on the App Store

Brain-Computer Interface Decodes Imagined Speech in Real Time

The Cell paper unveils a password-gated implant that deciphers inner speech in real time, raising immediate ethical as well as privacy concerns.

© Emory BrainGate Team
Image
Image
Image

Overview

  • Researchers used microelectrode arrays in the motor cortex of four participants with paralysis to decode inner speech from a 125,000-word vocabulary with up to 74% accuracy.
  • Participants achieved comfortable conversational rates of 120–150 words per minute by simply thinking sentences, marking a user-experience shift from previous attempted-speech BCIs.
  • The system employs a mental password—“chitty chitty bang bang”—detected with about 98% accuracy to unlock decoding and safeguard internal thoughts.
  • Current limitations include a small sample size, weaker neural activation for imagined speech compared with attempted speech, and inability to decode free-form spontaneous thought reliably.
  • Experts call for follow-up studies, hardware and algorithm improvements, larger clinical trials, and robust regulatory frameworks to address mental-privacy and ethical safeguards.