According to a recent UTS article, researchers at the GrapheneX-UTS Human-centric Artificial Intelligence Centre, part of the University of Technology Sydney, have unveiled a groundbreaking portable, non-invasive system that translates silent thoughts into text. The technology, showcased at the NeurIPS conference, involves participants silently reading text passages while wearing a cap that records electrical brain activity through an EEG. An AI model named DeWave then segments the EEG wave, capturing specific characteristics and patterns from the brain to translate them into words and sentences.
Unlike previous methods requiring surgery or MRI scans, this approach is both versatile and adaptable, potentially aiding communication for individuals unable to speak due to conditions like stroke or paralysis. The translation accuracy, currently at 40% on the BLEU-1 scale, is expected to improve, marking a significant breakthrough in neural decoding technology.
