Academic Emotion Prediction in Online Learning Utilizing Deep Learning Approach

As the world is progressing more towards new technology, more and more people are getting close to computers to perform their tasks. Computers have become an integral part of life. In recent years, web-based education has been perceived as a support tool for instructors as it gives the comfort of use at any time, and any place. In this situation, recognizing the user’s engagement with the system is important to make human-computer interaction more effective. Recognizing user engagement and emotions can play a crucial role in several applications including advertising, healthcare, autonomous vehicles, and e-learning. We focus on understanding the academic emotions of students during an online learning process. Four academic emotions namely, confusion, boredom, engagement, and frustration are considered here. Based on the academic emotions of students, we can incrementally improve the learning experience. In this paper, we have developed a system for identifying and monitoring the emotions of the scholar in an online learning platform and supplying personalized feedback to reinforce the online learning process.
 To achieve this, we have extracted images from the videos of the DAiSEE dataset and performed pre-processing steps like convert it into greyscale, detect a face from that image using OpenCV, change the size of the image, and then save it. Then labeling of the emotions is done and the model is trained using a convolution neural network (CNN) on the said images. In this way, the neural network is trained and can predict the emotion.