MUMBAI, India, Feb. 27 -- Intellectual Property India has published a patent application (202641019257 A) filed by Srm Institute Of Science And Technology; and Easwari Engineering College, Chennai, Tamil Nadu, on Feb. 19, for 'hybrid transformer-gnn framework for generalized mi based bci.'
Inventor(s) include Mr. R. Sudharsanan; Dr. B. Dwarakanath; Dr. P. Santhosh Kumar; Dr. K. Danesh; Mr. Vamsi Varun B; Mr. Saravanan S; and Mr. Poovarasan S.
The application for the patent was published on Feb. 27, under issue no. 09/2026.
According to the abstract released by the Intellectual Property India: "Motor Imagery-based Brain-Computer Interface systems let people move things that aren't in their bodies by sending signals from their brains. These Motor Imagery-based Brain-Computer Interface systems typically employ straightforward deep learning techniques such as CNNs and RNNs for computer instruction. These methods work well when the conditions are just right, but they have a lot of problems. For example, brain-computer interface systems that use motor imagery aren't very helpful for people. They also know a lot about how different people's brains work. Setting up Motor Imagery-based Brain-Computer Interface systems takes a long time for each person. Many of the systems we have now don't do a good job of showing how EEG signals change over time or how EEG electrodes are arranged in space. This makes them less helpful and less accurate. These issues make it harder for people to use MI-BCI systems in real life, which makes the problem of BCI illiteracy even worse. The new system uses a Transformer and a Graph Neural Network to help sort brain signals and fix these problems. The Transformer's job is to look at brain signals over time and find patterns. The Graph Neural Network part looks at how the electrodes are connected to see how the brain's signals move. The system also uses meta-learning, which is a way of learning that helps it quickly get used to new users, even if they haven't had much training. To test the system and make sure it can handle different subjects, we use benchmark datasets like BCI Competition IV-2a and PhysioNet EEG-MMI. The proposed method aims to enhance existing systems by optimizing their real-time performance, increasing their classification accuracy, and reducing the calibration duration. This project is working on making a MI-BCI system that is useful, scalable, and easy to use in the real world by combining unified spatial-temporal learning with user-adaptive mechanisms."
Disclaimer: Curated by HT Syndication.
