Automated emotion recognition based on facial images

Emotion recognition analysis is an interesting problem and it’s usually plays a vital role in our daily life. As computers have become an integral part of our lives, the interaction between human and computer become extremely important. Therefore, it’s essential for computer to identifying the situ...

Full description

Saved in:
Bibliographic Details
Main Author: Ang, Kah Chun
Other Authors: Teoh Eam Khwang
Format: Final Year Project
Language:English
Published: 2013
Subjects:
Online Access:http://hdl.handle.net/10356/52636
Tags: Add Tag
No Tags, Be the first to tag this record!
Institution: Nanyang Technological University
Language: English
Description
Summary:Emotion recognition analysis is an interesting problem and it’s usually plays a vital role in our daily life. As computers have become an integral part of our lives, the interaction between human and computer become extremely important. Therefore, it’s essential for computer to identifying the situation in which human do and respond upon it accordingly. Due subtle and complex of human emotion, it makes it very tough for achieving automated emotion recognition. Nowadays there are countless emotions expressions exist. This project aims to detect 6 elementary emotions: happy, sad, fear, disgust, anger and surprise. In order to achieve highly effective emotion recognition, this project aims to find the best feature extraction and classifier for the 6 basic types of emotions. Local binary pattern (LBP) and Histogram of Oriented Gradient are chosen to use for the feature extraction. LBP is used due to its simplicity in computation and useful in representing the texture. On the other hand, HOG is a newly method to use for face detection in the recent year. Thus, HOG is included in this project in compare with the result that obtained by using LBP. Support Vector Machine (SVM) is chosen as the feature classifier. The accuracy of the classifier by using two different feature extraction tools will be observe, compare and comment. This project is carried out by using Cohn-Kanade database. Throughout the experiment, Uniform LBP has turned out to be best in compare with other feature extraction methods that used in this project.