In a groundbreaking development, researchers at the GrapheneX-UTS Human-centric Artificial Intelligence Centre at the University of Technology Sydney (UTS) have successfully created a portable and non-invasive system that can translate thoughts into text. This revolutionary technology has the potential to greatly benefit individuals who are unable to speak due to conditions like stroke or paralysis. Additionally, it could enable seamless communication between humans and machines, such as the operation of bionic arms or robots.
The research conducted by the GrapheneX-UTS team has gained significant recognition, with their study being selected as a spotlight paper at the prestigious NeurIPS conference held in New Orleans. The team, led by Distinguished Professor CT Lin and supported by first author Yiqun Duan and PhD candidate Jinzhou Zhou from the UTS Faculty of Engineering and IT, aims to advance the field of artificial intelligence and machine learning.
During the study, participants wore a cap that recorded their electrical brain activity through an electroencephalogram (EEG) while silently reading passages of text. The EEG wave data collected was then segmented using an AI model named DeWave, developed by the researchers. DeWave leverages large quantities of EEG data to translate the brain signals into words and sentences.
According to Distinguished Professor Lin, this research represents a significant breakthrough in the field of neural decoding. Unlike previous methods that required invasive surgery or the use of large and expensive machinery, the GrapheneX-UTS system is non-invasive and portable. It is capable of translating brain signals into text without the need for additional aids like eye-tracking, making it more practical for real-life applications.
The UTS study sets itself apart from previous decoding technologies by its robustness and adaptability. Where earlier research relied on testing only one or two individuals, the team at GrapheneX-UTS included 29 participants. As EEG waves differ between individuals, this broader sample paves the way for more reliable and widely applicable results.
While the translation accuracy score of the current model hovers around 40% on BLEU-1, the researchers aspire to improve this metric to a level comparable to traditional language translation or speech recognition programs, which typically perform at around 90%. Although the model demonstrates a higher proficiency in matching verbs than nouns, it is still capable of forming meaningful sentences and aligning keywords.
The potential applications of this technology are vast and varied. It holds considerable promise in assisting individuals with speech impairments due to illness or injury. Additionally, it could revolutionize human-machine interactions, allowing for seamless communication and control of devices like bionic arms or robots.
The GrapheneX-UTS research builds upon previous advancements in brain-computer interface technology. Collaborating with the Australian Defence Force, UTS previously developed a system that uses brainwaves to command a quadruped robot. These developments highlight the university’s commitment to pushing boundaries and driving innovation in the field of artificial intelligence.
As the technology continues to evolve, the breakthrough achieved by the GrapheneX-UTS team brings us closer to a future where thoughts can be translated into text effortlessly. While there is still progress to be made, the implications of this research are incredibly promising and offer hope for improving communication and enhancing the quality of life for many individuals worldwide.
