Content area
Full Text
Introduction
Facial emotion recognition is defined as the process of anticipating bodily emotions such as facial expressions or brain impulses, and it is distinguished by qualities that allow us to distinguish one emotion from the others (Buono et al. 2022). As a result, the current research focuses on the automated identification of facial expressions in online learners. In a real-time scenario, to determine the diverse emotions, the learners' facial expressions are evaluated on the e-learning platform. Facial expressions are the movements of physical muscles caused by emotional impulses, including lip curling, brow raising, or brow wrinkling. A lot of information regarding online learners' emotional states may be determined by automatically watching the change in facial expressions (Savchenko and Makarov 2022).
The face intuitively contains a vast number of emotional information and has been directly connected to one's perceived involvement (Hassouneh et al. 2020). Recently, the study of physiological data acquired by sensors such as galvanic skin reaction, heart rate, and electroencephalogram (EEG) has been examined for assessing engagement. These sensors, however, were more obtrusive since users had to physically wear them, and they were also more costly than a camera (Ngai et al. 2022). Despite several research efforts for Facial Expression Recognition (FER) throughout the last decades, identifying learners' emotional states using facial expressions remained a difficult challenge in practical applications. Facial expressions come in a variety of intensities, ranging from modest micro-expressions to dramatic emotions (Hung et al. 2019) (Ninaus et al. 2019). The existing system detects the perceived and non-perceived behavior states which makes online learners more frustrated and causes dropout rates in online courses (Dewan, et al. 2018).
AI techniques including ML techniques and deep architectures prove their betterment in distinguishing the aspects based on the proper training that is employed to predict the students' mood at the present state (Mehta et al. 2022) (Ashwin and Guddeti 2020a). SVM, KNN, DT learning, association rule learning, rule-based ML, and others are the classic ML techniques that categorize the provided input data using handcrafted/predefined features (Zhang et al. 2020). However, the amount of processing power required for an NN is determined by the size of your data as well as the depth and complexity of the network. CNN design makes extensive use of computer...