Channel Selection of EEG Emotion Recognition using Stepwise Discriminant Analysis

EEG has been used by many applications recently, not only in the field of medicine but also telemarketing, games, and cybernetics. Measuring brain signal by involving EEG is complicated and delicate work because it involves many channels, frequency bands, and features. An efficient and effective met...

Full description

Saved in:
Bibliographic Details
Published in2018 International Conference on Computer Engineering, Network and Intelligent Multimedia (CENIM) pp. 14 - 19
Main Authors Pane, Evi Septiana, Wibawa, Adhi Dharma, Pumomo, Mauridhi Hery
Format Conference Proceeding
LanguageEnglish
Published IEEE 01.11.2018
Subjects
brain signal
Covariance matrices
electrode selection
Electroencephalography
Emotion recognition
Entropy
Feature extraction
Frequency measurement
identifying emotion
Signal processing algorithms
Wilks lambda
Online AccessGet full text
DOI10.1109/CENIM.2018.8711196

Cover

More Information
Summary:EEG has been used by many applications recently, not only in the field of medicine but also telemarketing, games, and cybernetics. Measuring brain signal by involving EEG is complicated and delicate work because it involves many channels, frequency bands, and features. An efficient and effective method in EEG measurement is then becoming crucial among the scientists. This paper proposed a channel selection study for emotion recognition based on the EEG signal by using Stepwise Discriminant Analysis (SDA). SDA is the extension of statistical tool for discriminant analysis that include stepwise technique. In this paper, the data was obtained from the public emotion EEG dataset which was recorded using 62 channels of EEG devices for three target emotions (i.e., positive, negative and neutral). In order to handle high dimensionality in EEG signals, we extracted differential entropy feature from five frequency bands: delta, theta, alpha, beta, and gamma. The selection criteria in SDA was based on Wilks Lambda score to get the optimal channel. In order to measure the performance of selected channels, we fed the features vector of the EEG signal to the LDA classifier. We conducted several scenarios from the different number of selected channels in experiments, such as 3, 4, 7, and 15 channels. The highest accuracy of 99.85% was obtained from 15 channels scenario in all combinations of frequency bands. Our results also confirm that alpha, beta, and gamma frequency bands are reliable for EEG emotion recognition.
DOI:10.1109/CENIM.2018.8711196