Two Emirati graduates from a university in Dubai have created a revolutionary AI solution aimed at assisting individuals with verbal impairments, such as those affected by stroke or paralysis, to communicate more effectively. Their innovative project utilizes an avatar-based artificial intelligence (AI) model that acts as a digital voice for users, converting their intentions into clear and expressive speech. This AI avatar serves as a communication bridge, enabling users to interact more naturally and inclusively. By offering a personalized, user-friendly experience, it has the potential to revolutionize assistive technology in healthcare, opening new avenues for inclusion and accessibility.
The creators of this technology, electrical engineering graduates Abdallah Al-Serkal and Ahmad Yousef, spent two years developing and refining this technology during their undergraduate studies. Al-Serkal, now a Master’s student at the Rochester Institute of Technology (RIT) Dubai, explained in an interview with Khaleej Times how the technology captures brain and eye signals through electro-encephalography (EEG) and electro-oculography (EOG) technology. “We use these methods to translate brain signals and eye movements into language, providing a new way for people with speech-related disabilities to communicate,” he said. He further explained, “EEG can detect brain biosignals. By integrating EEG with AI, we identified how the brain was thinking differently, when for example, a person is thinking of the numbers zero or one. We identified how certain thought patterns correspond to specific numbers or ideas. This concept began as a simple project and evolved over a summer of development.”
The words are articulated through an interactive avatar, which dynamically delivers the user’s thoughts through textual or visual expression. Al-Serkal shared that the project originally started as a lab assignment, which they expanded under the encouragement of their professor. The aim was to address communication barriers faced by stroke survivors and patients suffering from paralysis to enhance their quality of life. The prototype development involved data collection from 50 participants of different ages and backgrounds. “We captured the micro neural and eye movements that occur when participants think about different words. The test vocabulary focused on words related to patient needs, such as help, pain, medication, and sleep. Using machine learning, we translated these biosignals into words and validated the results, which were then conveyed through our AI avatar.” However, before a patient can use it efficiently, the individual needs to undergo training where the patient needs to view and think about specific words.
Currently, the AI can accurately interpret 10 specific words after a patient undergoes a brief training phase, where they think about each word while wearing small metal discs called electrodes attached to the scalp during EEG. “Once we collect this data, we label and feed it into the AI model. After training, the AI can predict which words the person is thinking of. The accuracy rate achieved with the prototype averaged 90 per cent overall, which was higher than we had anticipated,” Al Serkal noted.
Meanwhile, the drive to support public health and wellbeing has motivated the university to develop a micromaster program in health informatics and health management, which will launch soon. Associate Professors of Electrical Engineering, Dr Jinane Mounsef at RIT said, “It is vital to work on projects, in the university that have the potential to enhance wellbeing, such as this one focused on using AI to interpret brain signals and help patients communicate essential needs like hunger or help.” He added, “We foster an environment where technical innovation meets academic rigour, which empowers students to apply their knowledge for real-world impact. This project exemplifies how we strive to combine theory and practice to create solutions that improve the human condition.”
Source link: https://www.khaleejtimes.com
