Development of an automated online proctoring system

Abstract

The rapid development of machine learning technologies, the increasing availability of devices and widespread access to the Internet have significantly contributed to the growth of distance learning. Alongside distance learning systems, proctoring systems have emerged to assess student performance by simulating the work of a teacher. However, despite the development of image processing and machine learning technologies, modern proctoring systems still have limited functionality: some systems have not implemented computer vision methods and algorithms satisfactorily enough (false positives when working with students of different ancestry, racial background and nationalities) and classification of student actions (very strict requirements for student behaviour), so that some software products have even refused to use modules that use elements of artificial intelligence. It is also a problem that current systems are mainly focused on tracking students' faces and gaze and do not track their postures, actions, and emotional state. However, it is the assessment of actions and emotional state that is crucial not only for the learning process itself, but also for the well-being of students, as they spend long periods of time at computers or other devices during distance learning, which has a great impact on both their physical health and stress levels. Currently, control over these indicators lies solely with teachers or even students themselves, who have to work through test materials and independent work on their own. An additional problem is the quality of processing and storage of students' personal data, as most systems require students to be identified using their identity documents and store full, unanonymised video of students' work on their servers. Based on the analysis of all these problems that impede the learning process and potentially affect students' health in the long run, this article presents additional functional requirements for modern automated online proctoring systems, including the need to analyse human actions to assess physical activity and monitor hygiene practices when using computers in the learning process, as well as requirements for maximum protection of students' personal data. A prototype of the main components of an automated online proctoring system that meets the proposed requirements has been developed.