Engagement detection in e-learning environments using convolutional neural networks

Research output: Chapter in Book/Report/Conference proceedingPublished Conference contributionpeer-review

34 Citations (Scopus)

Abstract

Learning institutions are responsible for ensuring effective learning environments for online learners. An effective learning environment engages learners in educational activities. In this paper, we investigate the suitability of using three different popular models and a proposed model of convolutional neural network (CNN) for the engagement detection of online learners in their educational activities. By using the models, engagement detection has been done in a scalable and accessible manner by using facial expressions. The three models include all convolutional network (All-CNN), network-in-network (NiN-CNN), and very deep convolutional network (VD-CNN). These models are popular as they have simple network architectures and these have shown efficiency in different pattern classification applications. Each model has some unique features that made it different from the traditional as well as the other architectures. The All-CNN model replaces the max-pooling layer by a convolutional layer with increased stride. The NiN-CNN model replaces the linear convolutional layer with a multilayer perceptron and the fully connected layer by the same number of activation maps of the target classes. The VD-CNN model increases the depth by using small (3×3) convolutional filters. In the proposed model, we accumulate a number of advantageous features from the above three base models and achieved promising improvement. In the proposed model, the linear convolutional layer is replaced with a multilayer perceptron, the depth of the network is increased by using small (3×3) convolutional filters, and some max-pooling layers are replaced by convolutional layer with an increased stride. The three base models and the proposed model are applied on Dataset for the Affective States in E-Environments (DAiSEE) and analyzed their performances in a learner's engagement detection, where the proposed model outperforms the others.

Original languageEnglish
Title of host publicationProceedings - IEEE 17th International Conference on Dependable, Autonomic and Secure Computing, IEEE 17th International Conference on Pervasive Intelligence and Computing, IEEE 5th International Conference on Cloud and Big Data Computing, 4th Cyber Science and Technology Congress, DASC-PiCom-CBDCom-CyberSciTech 2019
Pages80-86
Number of pages7
ISBN (Electronic)9781728130248
DOIs
Publication statusPublished - Aug. 2019
Event17th IEEE International Conference on Dependable, Autonomic and Secure Computing, IEEE 17th International Conference on Pervasive Intelligence and Computing, IEEE 5th International Conference on Cloud and Big Data Computing, 4th Cyber Science and Technology Congress, DASC-PiCom-CBDCom-CyberSciTech 2019 - Fukuoka, Japan
Duration: 5 Aug. 20198 Aug. 2019

Publication series

NameProceedings - IEEE 17th International Conference on Dependable, Autonomic and Secure Computing, IEEE 17th International Conference on Pervasive Intelligence and Computing, IEEE 5th International Conference on Cloud and Big Data Computing, 4th Cyber Science and Technology Congress, DASC-PiCom-CBDCom-CyberSciTech 2019

Conference

Conference17th IEEE International Conference on Dependable, Autonomic and Secure Computing, IEEE 17th International Conference on Pervasive Intelligence and Computing, IEEE 5th International Conference on Cloud and Big Data Computing, 4th Cyber Science and Technology Congress, DASC-PiCom-CBDCom-CyberSciTech 2019
Country/TerritoryJapan
CityFukuoka
Period5/08/198/08/19

Keywords

  • Convolutional neural network
  • Deep learning
  • Engagement detection
  • Facial expressions
  • Online learning environment

Fingerprint

Dive into the research topics of 'Engagement detection in e-learning environments using convolutional neural networks'. Together they form a unique fingerprint.

Cite this