Emotion recognition based on spatially smooth spectral features of the EEG

The primary aim of this study was to select the optimal feature subset for discrimination of three dimensions of emotions (arousal, valence, liking) from subjects using electroencephalogram (EEG) signals. The EEG signals were collected from 25 channels on 21 healthy subjects whilst they were watching movie segments with emotional content. The band power values extracted from eleven frequency bands, namely delta (0.5-3.5 Hz), theta (4-7.5 Hz), alpha (8-12 Hz), beta (13-30 Hz), gamma (30-50 Hz), low theta (4-6 Hz), high theta (6-8 Hz), low alpha (8-10 Hz), high alpha (10-12 Hz), low beta (13-18 Hz) and high beta (18-30 Hz) bands, were used as EEG features. The most discriminative features for classification of EEG feature sets were selected using sequential floating forward search (SFFS) algorithm and a modified version of SFFS algorithm, which imposes the topographical smoothness of spectral features, along with linear discriminant analysis (LDA) classifier. The best classification accuracies for three emotional dimensions were obtained for liking (72.22%) followed by arousal (67.50%) and valence (66.67%). SFFS-LDA and modified SFFS-LDA algorithms produced slightly different classification accuracies. However, the findings suggested that the use of modified SFFS-LDA algorithm provides more robust feature subsets for understanding of underlying functional neuroanatomic mechanisms corresponding to distinct emotional states.