MUMBAI, India, Aug. 12 -- Intellectual Property India has published a patent application (202611075199 A) filed by K. R. Mangalam University on June 17, 2026, for A Lightweight System And Method For Cpu-Based Multimodal Retrieval Of Educational Video Segments.
Inventor includes Alina Banerjee.
The application for the patent was published on August 07, 2026, under issue no. 32/2026.
Abstract: A lightweight system (100) and method for retrieving segments of educational video are disclosed. The system (100) comprises a central processing unit (102) and a memory (104), and provides a lightweight feature-extraction subsystem (120) that segments a video into informative keyframes, extracts visual, textual and audio features, reduces the features to a 128-to-256-dimensional embedding, and fuses them by a weighted combination. The embeddings are stored with timestamps in a feature meta-database (130) and indexed by a central-processing-unit-efficient approximate nearest-neighbour index. A similarity matching module (140), a ranking module (150) and a video segment loader (160) return top-ranked segments with timestamps for a text, image, voice or concept query. The system (100) operates upon a central processing unit without a graphics processing unit, thereby reducing computational latency, memory and storage while enabling segment-level retrieval of educational content.
Disclaimer: Curated by HT Syndication.
