Modeling blended emotions in spontaneous Filipino laughter through facial expression analysis

Date of Publication

2014

Document Type

Master's Thesis

Degree Name

Master of Science in Computer Science

College

College of Computer Studies

Department/Unit

Computer Science

Thesis Adviser

Jocelynn W. Cu

Abstract/Summary

Laughter is a social signal commonly associated with positive emotions. However, there are cases in which it serves as a tool to conceal another emotion, resulting in a blend of the original and masked emotions. This research focused on recognizing the occurrence of blended emotions specifically in Filipino laughter by observing facial expressions. It is found out that by dividing the face into subregions, namely, the upper facial region and the lower facial region, blended emotions become discernible. To explore this concept, relevant facial point distances were extracted from the data. After several tests Multilayer Perceptron and Support Vector Machine proved to be the reliable algorithms to classify blended emotions. MLP yielded accuracy rates of 92.59%, 91.45%, and 90.79% for the whole, lower, and upper facial regions respectively. SVM yielded 93.50%, 88.81%, and 91.46% accuracy rates. Furthermore, it was discovered that by making a model of the combined lower and upper facial regions with labels, recognition rates significantly increased. Both algorithms now yielded accuracy rates of at least 95%.

Abstract Format

html

Language

English

Format

Electronic

Accession Number

CDTG005551

Shelf Location

Archives, The Learning Commons, 12F Henry Sy Sr. Hall

Physical Description

leaves ; 4 3/4 in.

This document is currently not available here.

Share

COinS