An Intelligent Examination Monitoring System Based on Deep Learning

Abstract

As Artificial Intelligence (AI) and education have merged to provide people with the opportunity to acquire new skills, the smart education system has exploded over the past decade. As the demand for smart proctoring services increases, AI-assisted proctoring solutions are in high demand. By developing a multimodal system, we eliminate the need for a human examiner to be available during the examination. Our system utilized a High definition (HD) camera and live window capture to obtain images. First, all humans in the image are identified, and then the head's feature points are calculated to determine its distance from other human heads. Facial feature analysis is performed to infer the candidate's expression. The immediate environment of the examinee can be taken up, including a cell phone, a piece of paper, or the mere existence of another person. Furthermore, our system monitors the examinee's mouth opening and facial deception. The combination of such models produces a smart rule-based prediction system that can figure out the probability that there was examination deception. We conducted a thorough evaluation of our system and determined that it performed well, with a precision of 0.95.