In a groundbreaking achievement, scientists from the GrapheneX-UTS Human-centric Artificial Intelligence Centre at the University of Technology Sydney (UTS) have introduced an innovative portable system capable of translating silent thoughts into text.
The technology, spearheaded by Distinguished Professor CT Lin and his team, is a remarkable leap in artificial intelligence and machine learning. The study, now a spotlight paper at the prestigious NeurIPS conference, showcases the system’s capability to facilitate communication for those facing speech challenges.
Participants in the study wore a cap recording electrical brain activity through an electroencephalogram (EEG) as they silently read passages. The EEG wave, analyzed by an AI model named DeWave, was segmented into distinct units, capturing unique characteristics from the participants’ brain patterns. This marks a significant breakthrough, eliminating the need for invasive procedures like brain electrode implants or restrictive MRI scans.
Professor Lin emphasized that “This research represents a pioneering effort in translating raw EEG waves directly into language, marking a significant breakthrough in the field.” Incorporating discrete encoding techniques and integrating large language models open new frontiers in neuroscience and AI.
Unlike previous methods that struggled to transform brain signals into words, the UTS technology, adaptable for use with or without eye-tracking, has demonstrated state-of-the-art performance. The study uses EEG signals from a cap rather than implanted electrodes to ensure a more robust and adaptable system.
The translation accuracy, currently at 40% on BLEU-1, is a notable achievement, surpassing previous benchmarks. While the model excels in matching verbs, challenges arise with nouns, showing a tendency toward synonymous pairs. Despite this, the model yields meaningful results, aligning keywords and forming similar sentence structures.
Looking ahead, the researchers aim to enhance the translation accuracy score to a level comparable to traditional language translation or speech recognition programs, reaching closer to 90%. This revolutionary technology builds upon UTS’s earlier achievements in brain-computer interface technology, showcasing the university’s commitment to advancing human-centric artificial intelligence.
This latest breakthrough follows UTS’s collaboration with the Australian Defence Force in developing brainwave-controlled robotics, highlighting the institution’s diverse contributions to cutting-edge technology.
In summary, the UTS researchers’ brain-to-text technology represents a remarkable stride in artificial intelligence and neuroscience, offering hope for improved communication for those with speech impairments and setting new standards for brain-computer interface technology.