| The prominence of online engineering education has notably increased. Various factors contribute to its effectiveness and relevance, making it a valuable facilitator of continuous learning and professional development in the dynamically evolving engineering field. To tackle this challenge, we introduce an intelligent framework that combines deep learning methods for the real-time detection of students' emotions during online learning and the assessment of their mental states regarding the taught content. Our framework comprises three modules. The first module employs a novel lightweight machine learning method, known as convolutional neural network-random forest (CNN-RF), to efficiently identify students' basic emotions during online courses. The second module involves mapping these basic emotions to an education-aware state of mind based on the Plutchik wheel. The third module is a visualization dashboard that provides educators with insights into students' emotional dynamics, enabling the identification of learning difficulties with high precision and offering informed recommendations for improvements in course content and online teaching methods. The proposed model has been trained and tested on the FER-2013 dataset, which is an open-labeled dataset for emotion recognition but not dedicated to a learning context. In this paper, we propose further experiments on a real learning situation. To assess the results of the framework, two evaluation questionnaires are proposed. The alignment between the reported outcomes and the results obtained from administered questionnaires serves to affirm the potential efficacy of our proposed model as a valuable support tool for establishing a good adaptive learning strategy. |
*** Title, author list and abstract as seen in the Camera-Ready version of the paper that was provided to Conference Committee. Small changes that may have occurred during processing by Springer may not appear in this window.