Journal article
Emotion classification from electroencephalogram using fuzzy support vector machine
Neural Information Processing, Vol.10634, pp.455-462
2017
Abstract
Realization of human emotion classification from Electroencephalogram (EEG) has great potential. Various methods in machine learning have been applied for EEG emotion classification and among these techniques, Support Vector Machines (SVMs) has demonstrated that it can provide good classification results. Therefore, SVM has been used widely in Affective Brain-Computer Interfaces (aBCI). However, EEG signals are non-stationary and they normally associate with outliers and uncertainties, and these issues could affect the performance of SVM. This study proposes the use of Fuzzy Support Vector Machine (FSVM) to deal with these issues. A benchmark dataset, Database for Emotion Analysis using Physiological Signals (DEAP), was used for subject-dependence classification. The experimental results showed that FSVM could deal with uncertainties and outliers, and enhanced the accuracies of arousal, valence and dominance classifications when compared to the SVM. Moreover, it was found that when gamma band was used as a feature from the two channels gave the best performance in comparison to other bands.
Details
- Title
- Emotion classification from electroencephalogram using fuzzy support vector machine
- Authors/Creators
- A. Chatchinarat (Author/Creator)K.W. Wong (Author/Creator)C.C. Fung (Author/Creator)
- Publication Details
- Neural Information Processing, Vol.10634, pp.455-462
- Publisher
- Springer Verlag
- Identifiers
- 991005540287707891
- Copyright
- © 2017 Springer International Publishing AG
- Murdoch Affiliation
- School of Engineering and Information Technology
- Language
- English
- Resource Type
- Journal article
Metrics
34 Record Views