The Library
Signal processing and machine learning methods with applications in EEG-based emotion recognition
Tools
Piho, Laura (2019) Signal processing and machine learning methods with applications in EEG-based emotion recognition. PhD thesis, University of Warwick.
|
PDF
WRAP_Theses_Piho_2019.pdf - Submitted Version - Requires a PDF viewer. Download (6Mb) | Preview |
Official URL: http://webcat.warwick.ac.uk/record=b3494555~S15
Abstract
Automatic emotion recognition has become increasingly popular, with applications in marketing, advertising, e-learning, entertainment, and more. Currently, the majority of automated emotion recognition is performed using facial expressions, body language, and speech intonation patterns. In recent years, using brain signals has become increasing popular. Being able to understand and analyse brain signals is beneficial in many applications. The goal of this thesis is to develop an effective method for extracting and representing EEG signals associated with human emotions, and to develop a robust classifier using machine learning tools for emotion recognition.
The thesis aims to address the common problems related to the EEG-based emotion recognition datasets, including dealing with small sample sizes, low signal-to-noise-ratio and high dimensional data. The contributions of this thesis lie in the proposed subject-dependent and subject-independent EEG-based emotion recognition frameworks. These frameworks are shown to accurately perform two-class classification as well as multi-class classification. In addition, a novel mutual information based signal reduction algorithm is introduced, aiming to increase the accuracy of EEG-based emotion recognition when the duration of the recording due to chosen stimuli is long. Furthermore, Gaussian Process classification is introduced for the purpose of EEG-based emotion recognition. This classifier is combined with the subject-dependent and subject-independent emotion recognition schemes and is shown to increase the accuracy when compared to the previous commonly used classifiers.
By using publicly available EEG datasets, the proposed novel frameworks are evaluated and shown to improve the EEG-based emotion recognition when compared against state-of-the-art methods. In addition, different signal processing methods suitable for EEG-based emotion recognition are introduced, explored, and analysed. An in-depth comparison of different feature extraction, feature selection, and classification methods is given using the proposed subject-dependent and subject-independent emotion recognition schemes.
Item Type: | Thesis (PhD) | ||||
---|---|---|---|---|---|
Subjects: | Q Science > Q Science (General) T Technology > TK Electrical engineering. Electronics Nuclear engineering |
||||
Library of Congress Subject Headings (LCSH): | Signal processing, Machine learning, Emotion recognition, Electroencephalography | ||||
Official Date: | June 2019 | ||||
Dates: |
|
||||
Institution: | University of Warwick | ||||
Theses Department: | School of Engineering | ||||
Thesis Type: | PhD | ||||
Publication Status: | Unpublished | ||||
Supervisor(s)/Advisor: | Tjahjadi, Tardi | ||||
Sponsors: | University of Warwick. School of Engineering | ||||
Format of File: | |||||
Extent: | xvi, 172 leaves : illustrations (some colour) | ||||
Language: | eng |
Request changes or add full text files to a record
Repository staff actions (login required)
View Item |
Downloads
Downloads per month over past year