MUMBAI, India, Nov. 2 -- Intellectual Property India has published a patent application (202641125479 A) filed by Rns Institute Of Technology; Dr. Mahantesh K; A T Vinay Reddy; Sanjana H G; N S Sai Sharan; and Ankita Anil Chougule on October 21, 2026, for Neuro-Sync Beats: Transformer-Based Eeg Emotion Recognition Using Differential Entropy Features..
Inventors include Dr. Mahantesh K; N S Sai Sharan; Sanjana H G; A T Vinay Reddy; and Ankita Anil Chougule.
The application for the patent was published on October 30, 2026, under issue no. 44/2026.
Abstract: Abstract The present invention proposes an intelligent brain–computer interface system for accurate and real-time human emotion recognition using electroencephalography (EEG) signals, integrating Differential Entropy (DE)–based feature extraction with a Transformer-based deep learning architecture. The system addresses critical limitations of existing EEG emotion recognition methods, which predominantly rely on conventional convolutional or recurrent neural networks and fail to effectively model long-range temporal dependencies and cross-channel correlations inherent in multi-channel EEG data. In the proposed method, raw EEG signals acquired from multi-electrode wearable devices are pre-processed through noise filtering and temporal segmentation, followed by extraction of Differential Entropy features across standard frequency bands including delta, theta, alpha, beta, and gamma. These features are structured as sequential embeddings and processed by a Transformer encoder employing multi-head self-attention mechanisms to capture both spatial and temporal neural dynamics relevant to emotional states. The learned representations are subsequently classified into discrete emotional categories using fully connected layers. The invention further enables adaptive application integration, wherein the recognized emotional state is automatically mapped to personalized digital actions such as mood-aware music recommendation, mental health regulation, or assistive human–computer interaction without manual intervention. Experimental evaluations demonstrate superior accuracy, robustness, and generalization across sessions compared to traditional deep learning approaches. The proposed system is scalable, non-invasive, and suitable for deployment in healthcare monitoring, affective computing, telemedicine platforms, and wearable neurotechnology applications. The combination of Differential Entropy feature modelling with Transformer-based attention learning constitutes a novel and non-obvious approach, rendering the invention suitable for patent protection.
Disclaimer: Curated by HT Syndication.
