MUMBAI, India, Nov. 2 -- Intellectual Property India has published a patent application (202641127080 A) filed by Dr. Ahmed Mudassar; and S. A. Engineering College on October 26, 2026, for Invigileai: A Real-Time Multi-Modal Ai System For Identity Verification And Cheating Prevention In Online Exams And Interviews.
Inventor includes Dr. Ahmed Mudassar Ali.
The application for the patent was published on October 30, 2026, under issue no. 44/2026.
Abstract: ABSTRACT The proposed system is an AI-assisted multimodal monitoring framework designed to analyze video frames and audio signals for identifying potential risk indicators during online examinations and interviews. The framework supports preprocessing and feature extraction, including the extraction of facial representations and relevant audio-visual features. It incorporates configurable modality weights to calculate a consolidated risk score based on available and valid monitoring indicators. Based on predefined risk thresholds, the system generates time-stamped alerts, maintains an event log, and presents relevant monitoring information through a supervisor dashboard to facilitate timely review and decision-making. Furthermore, configurable privacy settings, access-control mechanisms, and data-retention policies are integrated to protect candidate information and support secure data management. The framework aims to enhance monitoring efficiency, support transparent supervision, and promote privacy-aware assessment, while allowing human supervisors to review alerts before making decisions. The numerical comparison of the proposed InvigileAI framework with traditional monitoring and video-only monitoring can be presented using key performance metrics. For example, traditional monitoring may achieve an identity verification accuracy of 85%, while video-only monitoring may achieve 92%, compared with an illustrative 96% for the proposed InvigileAI framework. Similarly, cheating detection precision may be 75%, 84%, and 93%, respectively, while cheating detection recall may be 70%, 80%, and 90%. The false positive rate may decrease from 20% in traditional monitoring and 14% in video-only monitoring to 7% in the proposed framework. Furthermore, the average alert generation time may be reduced from 8 seconds and 4 seconds to 2 seconds, respectively. Processing latency may decrease from 120 ms per frame in video-only monitoring to 80 ms per frame in InvigileAI. These illustrative comparisons suggest that multimodal analysis integrating audio and video signals, configurable modality weights, and threshold-based alerts could improve detection performance and monitoring efficiency. All numerical values are hypothetical examples and must be replaced with experimentally measured results before publication.
Disclaimer: Curated by HT Syndication.
