Emotion classification from non-invasively measured electroencephalographic (EEG) data has been a growing research topic because of its potential application to affective brain-computer interfaces (ABCI), such as brain-inspired multimedia interaction and clinical assessment. A crucial component in ABCI is to reliably and accurately characterize individuals' brain dynamics into distinct affective states by employing advanced methods of pattern recognition. This chapter explores principles for translating neuroscientific findings into a practical ABCI. It will cover not only an overview of state-of-the-art EEG-based emotion recognition techniques, but also the basic research exploring neurophysiological EEG dynamics associated with affective responses. Although previous studies have demonstrated the use of EEG spectral dynamics for emotion classification, most of them achieved high classification accuracy by using an affective framework involving all available channels as well as Emotion Recognition: A Pattern Analysis Approach, First Edition. Edited by Amit Konar and Aruna Chakraborty.
316TOWARD AFFECTIVE BRAIN-COMPUTER INTERFACE frequency bands. The issue of feature and electrode reduction/selection has not typically been a primary goal in ABCI research. However, this chapter aims at resolving EEG feature selection and electrode reduction issues by the generalization of subjectindependent feature/electrode set extraction techniques that we have proposed in our series of emotion classification studies [1][2][3][4][5][6][7]. Furthermore, this study addresses several practical issues and potential challenges for ABCIs as well. We believe a userfriendly EEG cap with a small number of electrodes can efficiently detect affective states, and therefore significantly promote practical ABCI applications in daily life.
INTRODUCTION
Brain-Computer InterfaceBrain-computer interfaces (BCIs) translate human intentions into control signals to establish a direct communication channel between the human brain and output devices [8]. During the past two decades, BCI technology has become a hot research topic in the areas of neuroscience, neural engineering, medicine, and rehabilitation [9,10]. To detect human brain activities in real time, researchers have used different neuroimaging modalities such as electroencephalogram (EEG), magenetoencephalogram (MEG), electrocorticogram (ECoG), functional magnetic resonance imaging (fMRI), and functional near infrared spectroscopy (fNIRS) to build practical BCI systems [8]. EEG is a measurement of the electrical potential at a particular point on the scalp relative to another point on the head. It is believed that the EEG detected at scalp electrodes is generated by patches of cortical neurons with synchronous activity. A synchronous patch of cortex creates an electrical dipole that alternates in polarity as synaptic depolarization of the apical dendrites alternates with the consequent depolarization of the neuronal cell body. Each end of the dipole (dendrites and cell body) automatically rep...