MUMBAI, India, Aug. 12 -- Intellectual Property India has published a patent application (202611070228 A) filed by Ims Engineering College, Ghaziabad on June 04, 2026, for A System For Edge-Based Emotion Recognition Using Lightweight Neural Models.
Inventors include Dr. Manoj Kumar Singh; Dr. Sonali Mathur; Mr. Vivek Jain; and Dr. Nizam Uddin Khan.
The application for the patent was published on August 07, 2026, under issue no. 32/2026.
Abstract: The present invention relates to a device for edge-based emotion recognition using lightweight neural models. The device comprises a sensor interface module, preprocessing module, feature extraction module, lightweight neural inference module, model management module, confidence evaluation module, privacy control module, and output control module arranged within an edge computing platform. The device acquires facial, voice, physiological, behavioural, or contextual input signals through local sensors and processes the signals locally without transmitting raw user data to a cloud server for primary classification. Compact emotion-relevant features are extracted and analysed by lightweight neural models optimized through compression, quantization, pruning, distillation, or low-power architecture selection. A model management module adaptively selects inference configurations according to battery level, processor load, memory availability, thermal condition, and signal quality. Multimodal fusion and confidence evaluation improve reliability, while privacy controls restrict storage and transmission of sensitive raw data. The invention provides low-latency, privacy-preserving, energy-efficient emotion recognition suitable for edge devices. Accompanied Drawing [FIGS. 1-2]
Disclaimer: Curated by HT Syndication.
