- Title
- Scene perception using machine pareidolia of facial expressions
- Creator
- Hong, Kenny
- Relation
- University of Newcastle Research Higher Degree Thesis
- Resource Type
- thesis
- Date
- 2013
- Description
- Research Doctorate - Doctor of Philosophy (PhD)
- Description
- The aim of this thesis is to pursue the question, ‘Can a machine describe the emotional response to a scene using perceived faces and facial expressions?’. To establish the relevance of this interdisciplinary thesis question literature research was conducted not only in Computer Science but also in the disciplines of Cognitive Science and Psychology. The findings from this research revealed that humans can produce an emotional response to a scene using the facial expressions of perceived faces and face-like patterns, and this is achieved at a conscious and subconscious level, as well as beyond our visual focus. These findings provide the foundations for addressing the thesis question in the realm of computer vision and machine learning. A new face model, a new face detection algorithm and a new machine learning technique based on Support Vector Machines (SVMs) was developed and an extensive experimental evaluation was conducted. The new machine learning technique called ‘Pairwise Adaptive SVM’ (aka, pa-SVM) uses a refined parameter selection process for training, and was tested using real world datasets. The result is an improved detection and classification of faces and face-like patterns, as well as their associated facial expressions. The outcome is a machine that is capable of describing the emotional response to a scene using pareidolia of faces and facial expressions.
- Subject
- scene perception; facial expressions; classification; support vector machines; machine learning; face detection
- Identifier
- http://hdl.handle.net/1959.13/1039549
- Identifier
- uon:13669
- Rights
- Copyright 2013 Kenny Hong
- Language
- eng
- Full Text
- Hits: 1245
- Visitors: 2216
- Downloads: 521
Thumbnail | File | Description | Size | Format | |||
---|---|---|---|---|---|---|---|
View Details Download | ATTACHMENT01 | Abstract | 541 KB | Adobe Acrobat PDF | View Details Download | ||
View Details Download | ATTACHMENT02 | Thesis | 7 MB | Adobe Acrobat PDF | View Details Download |