MUMBAI, India, July 7 -- Intellectual Property India has published a patent application (202641081054 A) filed by Malla Reddy Engineering College For Women Autonomous; Malla Reddy University; Malla Reddy Mr Deemed To Be University; and Malla Reddy Vishwavidyapeeth Deemed on July 01, 2026, for Facial-Feature Based Drowsiness Detection Using Mcnn.

Inventors include Dr. Y. Madhaveelatha; Mr. Raghuvaran Cheerla; Ms Siva Parvathi Gorla; Ms Madhavi Manasu; Mr. Ramesh Challa; Dr. B. Giridhar; Ms. Thota Anitha; and Mrs. Kotte Shivani.

The application for the patent was published on July 03, 2026, under issue no. 27/2026.

Abstract: The present invention discloses a Facial-Feature Based Drowsiness Detection system using a Multi-Scale Convolutional Neural Network (MCNN) framework, designed to prevent road accidents caused by driver drowsiness. Driver fatigue resulting from lack of sleep, extended driving hours, and health conditions is a leading cause of vehicular fatalities globally. Existing systems fail to provide accurate, real-time drowsiness classification. The present invention addresses these limitations through a robust deep learning pipeline integrating facial landmark detection, advanced image preprocessing, and hybrid feature extraction. The system utilizes YAWDD and NTHU-DDD benchmark datasets of video sequences of drivers under real-world conditions. Video frames are preprocessed using Cross Guided Bilateral Filtering. Facial landmark localization is performed using the Dlib library. Feature extraction employs a hybrid dual-tree complex wavelet transform combined with Walsh-Hadamard transform. Features are optimized using the Flamingo Search Algorithm (FSA) and classified using MCNN. Experimental results demonstrate that the proposed MCNN-FSA model achieves 98% accuracy on the YAWDD dataset, outperforming conventional methods including AlexNet, ResNet, VGG, Random Forest, Naive Bayes, KNN, and AdaBoost. The system is deployed as a web-based Flask application providing real-time drowsiness prediction with user authentication.

Disclaimer: Curated by HT Syndication.