USC Researchers Develop Non-Invasive Brain-to-Speech AI System

Jihwan Lee of USC's Signal Analysis and Interpretation Laboratory (SAIL) is pioneering non-invasive brain-to-speech technology using EEG, offering new hope for individuals with speech impairments. Unlike traditional invasive methods like ECoG, EEG allows brain signal decoding without surgery, making brain-computer interfaces more accessible.

Lee’s team has developed a novel dual-module system that simultaneously decodes phoneme sequences and audio waveforms, improving accuracy and reducing latency. Their recent study, Enhancing Listened Speech Decoding from EEG, showed the method is twice as effective as earlier models in removing noise and decoding speech from EEG signals.

Powered by advances in AI and biosensing, the system decodes speech from the intention to speak, helping those unable to vocalize. Senior researcher Shrikanth Narayanan highlighted its potential to transform communication for patients with ALS or paralysis. Lee also emphasized the importance of international collaboration, including partnerships with top Indian institutions, to expand the technology’s reach.

21-05-2025