MUMBAI, India, June 30 -- Intellectual Property India has published a patent application (202641074880 A) filed by Andhra University on June 17, 2026, for Federated Cross-Subject Electroencephalogram Emotion Recognition System With Similarity Guided Evidence Fusion.
Inventors include Bhanusri Andhavarapu; and Dr. Shashi Mogalla.
The application for the patent was published on June 26, 2026, under issue no. 26/2026.
Abstract: 049] The present invention relates to a federated cross-subject electroencephalogram emotion recognition system configured to recognize emotional states across different subjects without transferring raw electroencephalogram signals to a central server. The system includes a plurality of local subject devices that acquire electroencephalogram signals, preprocess the signals, generate latent representations, and compute emotion-specific local prototypes. A federated coordination server receives privacy-preserving prototype updates from the local subject devices and generates reference prototypes for respective emotion classes by using reliability, confidence, class availability, and signal quality information. During inference, a local subject device compares a target electroencephalogram representation with the reference prototypes to generate similarity evidence and combines the similarity evidence with classifier evidence, subject similarity evidence, signal quality evidence, and uncertainty evidence. The invention improves cross-subject emotion recognition accuracy, reduces subject-specific calibration, reduces communication payload, protects sensitive neurophysiological data, and improves reliability in heterogeneous electroencephalogram acquisition environments. Accompanied Drawing [FIGS. 1- 2]
Disclaimer: Curated by HT Syndication.
