Recognition of Academic Emotions in Online Classes

Abstract

Online education has proliferated since the COVID-19 pandemic. Classes have been moved online as a result of school closures. Despite the flexibility offered by online learning, there are several challenges faced. Creating a good classroom climate for online classes is a challenging task. It is difficult for the teachers to obtain emotional feedback from the students, especially in asynchronous classes or classes with large number of students. It is hard for the teachers to evaluate the engagement of the students in class without knowing the students’ emotional response. The existing facial expression recognition databases focus on basic human emotions like happy, angry, sad, surprise and neutral. These basic emotions are not appropriate for learning as psychological and pedagogical studies have shown that there are differences between basic human emotions and academic emotions. In view of these problems, this paper presents a study on academic emotions. A dataset comprising four pertinent academic emotions have been established. Empirical analysis on the dataset is conducted using both hand crafted and deep learning approaches. The baseline evaluation demonstrates the suitability of the established academic dataset for online learning.