MUMBAI, India, Sept. 22 -- Intellectual Property India has published a patent application (202611091871 A) filed by Maharishi Markandeshwar Deemed To Be University on July 29, 2026, for Rnn-Driven Abnormal Human Activity Detection For Intelligent Surveillance Systems.
Inventor includes Dr. Tejinder Kaur.
The application for the patent was published on September 18, 2026, under issue no. 38/2026.
Abstract: The present invention relates to an intelligent surveillance system and method for automatic detection and classification of abnormal human activities using a hybrid deep learning architecture. The system receives surveillance video sequences, extracts image frames, performs preprocessing including resizing and normalization, and generates fixed-length temporal sequences. A Convolutional Neural Network extracts spatial features representing body posture, motion regions, and scene context, while a Long Short-Term Memory network learns temporal relationships among consecutive frames for accurate activity recognition. A Softmax classifier identifies multiple activity categories including fighting, theft, assault, burglary, vandalism, robbery, road accidents, shooting, arson, shoplifting, abuse, arrest, explosion, and normal activity. Upon detection of an abnormal activity, the system generates real-time graphical alerts and automatically transmits electronic mail notifications to predetermined users. The system is deployable through a Streamlit-based web application and is trained using the Dynamic Activity Classification System (DACSS) dataset. The proposed invention improves surveillance efficiency by integrating spatial and temporal deep learning with automated notification capabilities for real-time intelligent monitoring and enhanced public safety.
Disclaimer: Curated by HT Syndication.
