EEG-based emotion classification using wavelet based features and support vector machine classifier
As technology and the understanding of emotions are evolving, there are numerous opportunities for classification of emotion due to the high demand in the psychophysiological research. The researches need an efficient mechanism to recognise the various emotions precisely with less computation comple...
Saved in:
| Main Author: | |
|---|---|
| Format: | Thesis |
| Language: | English |
| Published: |
2015
|
| Subjects: | |
| Online Access: | http://eprints.utm.my/id/eprint/53780/25/NormaslizaMRazaliMFC2015.pdf |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | As technology and the understanding of emotions are evolving, there are numerous opportunities for classification of emotion due to the high demand in the psychophysiological research. The researches need an efficient mechanism to recognise the various emotions precisely with less computation complexity. The current methods available are too complex with higher computational time. This study proposes a classification of human emotion using electroencephalogram signals (EEG). The study utilised electroencephalogram signals (EEG) to classify emotions which is positive/negative arousal, valence and normal emotions. Electroencephalogram signals (EEG) are analysed from 4 different participants from the dataset that acquire from the public data source. These dataset go through several processes before the derivation of the features such as preprocessing using band pass filtering and artifacts removals, segmentation of the signals and Multiwavelet Transform (MWT) analysis of the processed data. The signals are decomposed up to level 3 decomposition and detail coefficients are used for features extraction. Statistical and power spectral density (PSD) features are computed and feed into the classifiers. Simple classification methods Support Vector Machine (SVM) is used to classify the emotion and their performances are evaluated. The experimental results report that statistical features and Support Vector Machine (SVM) achieved better accuracy up to 75.8%, 72.3% and 74.0% for arousal, valence and normal class respectively. In conclusion this research suggests the use of Multiwavelet Analysis for future work on recognizing various emotions from the Electroencephalogram signals (EEG). |
|---|