Classification of Affective States and their Level in a Learning Environment using Neural Networks

Bajaj, Kishan Kumar (2022) Classification of Affective States and their Level in a Learning Environment using Neural Networks. Masters thesis, Dublin, National College of Ireland.

Preview	PDF (Master of Science) Download (6MB) | Preview
Preview	PDF (Configuration manual) Download (880kB) | Preview

Abstract

Online learning has become the way of life from the last two years due to the ongoing pandemic which has completely disrupted the classroom teaching experience and the implicit feedback loop between the teacher and the students. Like technology is being used to maintain continuity in learning, it can also be used to create a feedback mechanism between the teacher and student in an online learning environment. The key to develop this is detection of different affective states and their levels exhibited by the students during learning. In this research, five different models are presented; one model to classify the correct out of four affective states (boredom, engagement, confusion and frustration) and for each affective state a separate model to classify the correct level out of four levels of each affective state (very low, low, high, very high). All these models are built upon a hybrid ResNet + TCN neural network architecture and are trained using publicly available DAiSEE dataset. The data set contains videos which are converted to sequence of frames for training the model. Another dataset taken from EmotiW2020 Challenge is also used to cross validate the engagement level classification model. The affective state classification model and the boredom level classification model outperform existing works. Confusion and Frustration level models perform almost at par with the existing models. Engagement level classification model performs at par with other models but considering the fractional amount of training data and iterations used, the performance can be considered as good as the existing baseline models. This model is assessed on, both, DAiSEE and EmotiW2020 data sets and achieves similar performance on both data sets.

Item Type:	Thesis (Masters)
Subjects:	Q Science > QA Mathematics > Electronic computers. Computer science T Technology > T Technology (General) > Information Technology > Electronic computers. Computer science L Education > LC Special aspects / Types of education > Blended Learning L Education > LC Special aspects / Types of education > E-Learning
Divisions:	School of Computing > Master of Science in Data Analytics
Depositing User:	Tamara Malone
Date Deposited:	18 Jan 2023 15:18
Last Modified:	18 Jan 2023 15:18
URI:	https://norma.ncirl.ie/id/eprint/6079

Actions (login required)

View Item