MUMBAI, India, June 30 -- Intellectual Property India has published a patent application (202641054001 A) filed by Sini Prabhakar; Sushma S; Saranyasree R; Safrin Farjana M; and Vishnupriya N on April 28, 2026, for Examora.
Inventors include Sini Prabhakar; Sushma S; Saranyasree R; Safrin Farjana M; and Vishnupriya N.
The application for the patent was published on June 26, 2026, under issue no. 26/2026.
Abstract: ABSTRACT EXAMORA is an intelligent Al-driven online examination system that integrates two primary capabilities: malpractice detection and automated question generation. The malpractice detection module leverages deep learning to monitor student behavior during remote examinations and detect suspicious activities such as impersonation, gaze deviation, tab switching, and audio anomalies. Rather than making real-time accusations, the system computes a cumulative risk score after exam completion, reducing false positives and ensuring ethical decision-making. The automated question generation module employs a Retrieval-Augmented Generation (RAG) pipeline, allowing administrators to upload domain-specific documents and automatically generate structured, mark-weighted examination papers. The system retrieves contextually relevant content from uploaded documents and uses Mistral 7B, a locally hosted language model via Ollama, to synthesize questions calibrated to specified difficulty levels and marking schemes. Developed using Python and modern Al frameworks, EXAMORA integrates computer vision, temporal behavior analysis, anomaly detection, vector-based document retrieval, and large language model inference. It uses FaceNet for identity verification and MediaPipe for facial landmark detection. EXAMORA preserves privacy by processing all data locally and storing only extracted features. The system also provides explainable results through feature analysis and anomaly timelines, while generating examination papers that are strictly grounded in uploaded content. Keywords: Online Proctoring, Malpractice Detection, Deep Learning, Face Recognition, Explainable Al, Privacy Preservation, Anomaly Detection, Retrieval-Augmented Generation, Question Generation, Large Language Model, Vector Database
Disclaimer: Curated by HT Syndication.
