Document Type

Dissertation

Rights

This item is available under a Creative Commons License for non-commercial use only

Disciplines

1.2 COMPUTER AND INFORMATION SCIENCE

Publication Details

A dissertation submitted in partial fulfilment of the requirements of Technological University Dublin for the degree of M.Sc. in Computer Science (Data Science) 2020.

Abstract

Multimodal Emotion Recognition is an emerging associative field in the area of Human Computer Interaction and Sentiment Analysis. It extracts information from each modality to predict the emotions accurately. In this research, Bimodal Emotion Recognition framework is developed with the decision-level fusion of Audio and Video modality using RAVDES dataset. Designing such frameworks are computationally expensive and require more time to train the network. Thus, a relatively small dataset has been used for the scope of this research. The conducted research is inspired by the use of neural networks for emotion classification from multimodal data. The developed framework further confirmed the fact that merging modality can enhance the accuracy in classifying emotions. Later, decision-level fusion is further explored with changes in the architecture of the Unimodal networks. The research showed that the Bimodal framework formed with the fusion of unimodal networks having wide layer with more nodes outperformed the framework designed with the fusion of narrow unimodal networks having lesser nodes.

DOI

https://doi.org/10.21427/gpbz-2319

Share

COinS