Detection of Emotions from Electroencephalographic Recordings by Means of a Nonlinear Functional Connectivity Measure

Abstract

The brain has been typically assessed as a group of independent structures focused on the realization of determined processes separately. Nevertheless, recent findings have confirmed the existence of interconnections between all brain regions, thus demonstrating that the brain works as a network. These areas can be interconnected either physically, by anatomical links, or functionally, through functional associations created for a coordinated development of mental tasks. In this sense, the assessment of functional connectivity is crucial for discovering new information about the brain’s behavior in different scenarios. In the present study, the nonlinear functional connectivity metric cross-sample entropy (CSE) is applied in the research field of emotions recognition from EEG recordings. Concretely, CSE is computed to discern between four different emotional states. The results obtained indicated that the strongest coordination appears in intra- and inter-hemispheric interactions of central, parietal and occipital brain regions, whereas associations between left frontal and temporal lobes with the rest of areas show the most dissimilar dynamics, thus a higher uncoordinated activity. In addition, coordination is globally higher under emotional conditions of high arousal/low valence (like fear or distress) and low arousal/high valence (such as relaxation or calmness).