A Framework for Facial Expression Recognition Based Feedback Tracking in Online Educational Platforms

Abstract

During times of worldwide pandemic, online teaching platforms are becoming increasingly popular as an alternative to traditional learning environment. However, while many online platforms have been reported in the literature, many of them lack a reliable feedback tracking mechanism through which the emotional state of students or users can be determined for effective teaching and learning. This has necessitated the need to design and simulate an online feedback system. Sample facial images were obtained from FER-2013 Dataset, and preprocessed. Principal Component Analysis (PCA) was used in extracting the facial features in the sample images. A total of 31,885 images with six different emotional classifications which are happy, sad, fear, neutral, angry, disgust were considered. The images were then split into training and test images with train consisting of 75% of the whole dataset while test had 25%. Ensemble of Support Vector Machine, Random Forest and k-nearest neighbour were used in classifying the images. The results of classification serve as inputs to the feedback tracking mechanism of the proposed model, which was formulated as an algorithm. The performance of the proposed ensemble model was compared with its based classifiers using metrics such as Precision, Recall, F1- score, and Accuracy. The simulation results during training showed that ensemble (SVM +RF+KNN) had accuracy 98.92%, RF had accuracy 92.94%, SVM had accuracy 95.57%., and KNN had accuracy 85.42%. Likewise in test dataset, Ensemble (SVM +RF+KNN) had accuracy 93.92%, RF had accuracy 90.94%, SVM had accuracy 91.57%, and KNN had accuracy 83.42%. Other performance metrics such as Precision, F1-Score, and Recall were also measured during simulation. The study designed a system for obtaining feedback based on facial expression of online participants for an improved online educational platform usage and acceptance. The developed online feedback framework could be integrated into the existing online educational platforms for improved usage and acceptance.