Enhancing E-Learning: EEG Signal Classification to Evaluate Students' Understanding of Online Lectures

Online learning is increasingly popular, and the COVID-19 pandemic has only accelerated its adoption. Neverthe-less, a significant challenge in online education is accurately assessing students' comprehension and engagement during lectures. Conventional evaluation methods like quizzes and assignments often fall short in providing a precise representation of students' understanding and lack real-time feedback. In our proposed approach, we tackle this challenge by developing a deep learning-based classifier capable of identifying students who have grasped the online lecture content and those who haven't. We sourced a 14-electrode EEG dataset from an online repository. During EEG signal recording, multiple frequencies, including noise, artifacts, and biological signals, were present. We addressed this issue by applying the Fast Fourier Transform to denoise the raw EEG sig-nals. To classify the EEG dataset using machine learning models, we employed two techniques: the Short-time Fourier Transform and Continuous Wavelet Transform for feature extraction. For each feature, we created two classes of images-one for students who comprehended the lecture and another for those who did not. In pursuit of this task, we utilized four pre-trained models, namely VGG-16, VGG-19, ResNet-50, and ResNet-101. Among these models, ResNet-50 emerged as the top performer, achieving an accuracy rate of 85.31% for Continuous Wavelet Transform, alongside an F1 score of 87%. This research presents a promising approach to assessing online learning comprehension in real-time, leveraging advanced deep learning techniques and EEG data analysis.