Eye Gaze Analysis of Students in Educational Systems

Eye gaze provides indicative information about the status and the behavior of a person and can be very assistive in human-computer interaction. Eye-gaze analysis is very helpful in a variety of applications in order to understand the interest of the users, their behavior or even to unveil distractions. However, the accurate eye-gaze estimation is a very challenging process. In this paper, we present an eye gaze estimation work that relies on convolutional neural networks which imitate the LeNet’s architecture. They analyze eye gaze and provide a 2D vector that concerns the coordinates of the specific pixel inside the 2D screen’s space, in which the user is looking at. Also, a system capable of working under various real-world conditions such as light, angle and distance differentiations was designed and developed. An evaluation study was performed and the results are quite promising pointing out that the system is scalable and accurate in estimating the eye gaze of the users.