Attention Detection using Electro-oculography Signals in E-learning Environment

Recently, on account of the Covid-19 pandemic online learning has become a main strategy of learning in any educational environment. The main problem is the lack of focus or attention during a student watching an online lecture or video due to ways of distraction that can occur and that's what interrupts the students and make them lose their focus. In this research paper, we propose the attention detection system using an Electro-oculogram (EOG) signal generated by eye movement. The first goal of this study was detecting the un-focusing periods in the video. The proposed process of recognizing the un-focusing period depends on recognizing the eye movement direction. 6 classes of eye movements were classified in this study: left, right, up, down, blinking, and no movement. The process started by collecting our benchmark dataset from 50 subjects, which was considered the largest dataset at all for eye movements, using a special hardware device. The pre-processing step was down to filter the signals from any noise that may be caused by the surrounding environment or involuntary movements using the band-pass filter. Finally, the classification step was down. Three models of deep learning were examined for the best classifier: Convolution Neural Network (CNN), Inception Network, and VGG Network. Our experiments achieved the best results with the Inception model, reaching an average accuracy of 93.63%. The second goal of this study was to recognize the Attention-deficit hyperactivity disorder (ADHD) pattern in the student’s recording. The model succeeded to recognize this disordered pattern.