According to research from the GrapheneX-UTS Human-centric Artificial Intelligence Centre at the University of Technology Sydney (UTS), a portable, non-invasive system is capable of decoding silent thoughts and transforming them into text.

In the research, participants silently read passages of text while wearing a cap that recorded electrical brain activity through their scalp, utilizing an electroencephalogram (EEG). The recorded EEG waves were then segmented into distinct units capturing specific characteristics and patterns from the human brain. 

This process was facilitated by an AI model called DeWave, developed by the researchers. DeWave translates EEG signals into words and sentences by learning from extensive EEG data.

“This research represents a pioneering effort in translating raw EEG waves directly into language, marking a significant breakthrough in the field,” said Professor CT Lin, director of the GrapheneX-UTS HAI Centre.

“It is the first to incorporate discrete encoding techniques in the brain-to-text translation process, introducing an innovative approach to neural decoding. The integration with large language models is also opening new frontiers in neuroscience and AI,” he continued. 

Notably, this technology surpasses previous methods that required surgical implantation of electrodes in the brain, such as Elon Musk’s Neuralink, or the use of large and expensive MRI machines. The new system can be used with or without eye-tracking, making it more practical for daily life.

Recommended reading

• NHS Scotland Signs New Cloud Deal for Diagnostic Imaging
• Edinburgh Startup Secures £150K for Its Brain Imaging Tech
• Edinburgh Startup Secures £150K for Its Brain Imaging Tech

“The model is more adept at matching verbs than nouns. However, when it comes to nouns, we saw a tendency towards synonymous pairs rather than precise translations, such as ‘the man’ instead of ‘the author,'” said Yiqun Dua, Ph D candidate at UTS.

“We think this is because when the brain processes these words, semantically similar words might produce similar brain wave patterns. Despite the challenges, our model yields meaningful results, aligning keywords and forming similar sentence structures,” he said.

The translation accuracy score, currently at 40% on BLEU-1, is expected to improve, aiming for levels comparable to traditional language translation or speech recognition programs, which typically reach around 90%.