EEG-based multi-frequency band functional connectivity analysis and the application of spatio-temporal features in emotion recognition

The study of emotional states in brain-computer interface (BCI) has a wide range of applications in psychiatry, psychology, et al. However, there is few novel feature extraction method integrating time-domain and space-domain features in emotion classification. This study explored the connectivity patterns between brain regions over functional connectivity brain networks in different frequency bands of electroencephalogram (EEG) signals and proposed a novel feature extraction method to classify emotions, which provided a unique perspective on emotion recognition. We constructed phase locking value (PLV) matrices analyzed in different frequency bands. Then, three distance matrices, dF, dS, and dLE, were built using the corresponding three distance measures (the Frobenius norm, the spectral norm, and the log-Euclidean distance, respectively). And the complexity measures on those distance matrices were calculated. The distance matrices and complexity measures, as two features, were fed into the machine learning classifiers to validate the proposed method. Eventually, the dF matrix obtained an average classification accuracy of 83.96 % in the alpha band between positive and neutral emotions, the dLE matrix obtained an average classification accuracy of 84.12 % in the beta band between positive and negative emotions, and the dF matrix obtained an average classification accuracy of 83.56 % in the delta band between neutral and negative emotions. We conclude that the delta, alpha, and beta frequency bands correlate highly with emotions, and the brain's anterior and right temporal lobes are inextricably linked to emotions. In addition, the feature extraction method proposed in this paper can effectively improve the classification accuracy of emotions.