Abstract:
Online education platforms, including massive open online courses (MOOCs), have spread in recent years. However, the proctoring processes used in them haven’t accompanied this recent growth and importance of this kind of learning. In this thesis, a biometric-based approach is developed to tackle this problem, it eliminates the need for a human proctor to be present as the proctoring process held automatically using examinee’s computer camera and microphone. Computer vision and audio analysis techniques are being used in order to determine the state of the examinee, record examinee’s activity throughout the examination and for detecting if there are any malpractices present. This approach provides continuous verification of the examinee which eliminates the ability for an unauthorized person to take the examination. In addition, it detects if the examinee is missing from the front of the camera or if another person is present beside the examinee. The speech detection is also developed to detect if the examinee speaks with someone else during the examination. This approach is feasible for large-scale use as it does not use any expensive hardware, as well as the detection of malpractices is done automatically which reduce the cost of the monitoring process. Experimentations proved that this approach has high accuracy in detecting various cheating behaviours and it has an advantage over currently available systems.
