Nonlinear methodologies applied to automatic recognition of emotions: An EEG review

Abstract

Development of algorithms for automatic detection of emotions is essential to improve affective skills of human-computer interfaces. In the literature, a wide variety of linear methodologies have been applied with the aim of defining the brain’s performance under different emotional states. Nevertheless, recent findings have demonstrated the nonlinear and dynamic behavior of the brain. Thus, the use of nonlinear analysis techniques has notably increased, reporting promising results with respect to traditional linear methods. In this sense, this work presents a review of the latest advances in the field, exploring the main nonlinear metrics used for emotion recognition from EEG recordings.