Approach on affective valence detection from EEG
signals based on global field power measure and
SVM-RFE algorithm
A. R. Hidalgo-Mu˜noz
a
, M. M. L´opez
b
, A. M. Tom´e
c
, A. T. Pereira
d
, and
I. M. Santos
d
a
Dept. Experimental Psychology, University of Seville, 41018 Seville, Spain
b
IEETA, University of Aveiro, 3810-193 Aveiro, Portugal
c
DETI/IEETA, University of Aveiro, 3810-193 Aveiro, Portugal
d
Dept. Ciˆencias de Educac¸˜ao, University of Aveiro, 3810-193 Aveiro, Portugal
Abstract. EEG signals have attracted the interest of scientific community for
understanding how brain processes emotions. In order to extract objective con-
clusions, automatized methods that are able to reinforce the subjective visual
explorations of the signals are desirable. In this work, a feature extraction +
wrapped classification scheme is proposed for analysing how brain reacts to vi-
sual high/low valence stimuli and how the linked brain processes change when a
novel or familiar stimulus is presented. For such purpose, experiments were car-
ried out using the international affective picture system (IAPS) images. Global
field power (GFP) from the recorded EEG signals is computed, and a support
vector machine-recursive feature elimination (SVM-RFE) method is applied to
the input signals. The combination of these techniques yielded up to 100% peak
accuracy in both classification tasks, outperforming traditional statistical meth-
ods for group comparisons such as t-test.
Keywords: Affective computing; EEG; Global field power; Habituation; SVM;
SVM-RFE
1 Introduction
The dimensional model of emotions asserts that emotions can be mainly defined by
two dimensions: arousal and affective valence [1]. Some studies concentrated on one
of the dimensions of the space like identifying the arousal intensity (high versus low)
or the valence (low/negative versus high/positive), and eventually a third class neutral
state. Normally, emotions are elicited by (i) presenting an external stimulus (picture,
sound, word or video) related to different emotions at some predefined interval, or by
(ii) simply asking subjects to imagine different kinds of emotions.
Affective valence is an interesting, essential dimension in studies about emotion
processing, which is influenced by several variables beyond the subject’s anatomy and
psychophysiology. Habituation is one of these variables, and consists of a reduction in
the response to a stimulus when it is repeatedly presented [2] and affects either emo-
tional or attentional processes [3, 4]. The interaction between habituation and affective
valence processing is affected in diverse psychological and psychiatric disorders, like
Granada, 18-20 March, 2013