Authors:
Yujun Niu
1
;
Hao Zhang
2
;
Shin'ichi Warisawa
1
and
Ichiro Yamada
1
Affiliations:
1
University of Tokyo, Japan
;
2
The University of Tokyo, Japan
Keyword(s):
Arousal Recognition, Electroencephalography (EEG), Discrete Wavelet Transform (DWT), Channel Selection.
Related
Ontology
Subjects/Areas/Topics:
Affective Computing
;
Biomedical Engineering
;
Evaluation and Use of Healthcare IT
;
Health Information Systems
;
ICT, Ageing and Disability
;
Pattern Recognition and Machine Learning
Abstract:
Improving arousal recognition accuracy by using EEG signals is important for emotion recognition. In this
research, discrete wavelet transform is used to extract features, and a cross-level method is adopted to select
effective features. The cross-level method shows great potential for two-level arousal classification, and the
recognition accuracy reaches 91.8%. The sensitivity of EEG channels is also discussed based on two
ranking methods of SCP (single-channel performance) and ANOVA (analysis of variance). Finally, arousal
recognition method based on EEG signals is applied to construct a Japanese emotion database.