The dynamics of emotions: a preliminary study on continuously annotated arousal signals

Human emotions are characterized by a complex interaction between conscious experience, physiological arousal, and social dimension. Although the importance of considering emotion response as a nonlinear dynamical system is widely recognized, mathematical models able to describe the time-varying conscious emotional states are still lacking. In recent literature on Affective Computing, novel annotating tools have been introduced to record continuous self-assessed emotion ratings. These data represent a valuable source to describe the dynamics arising during the conscious experience of emotions. Therefore, in this study, we investigate the trajectories traced in the reconstructed phase space of continuously annotated arousal signals acquired during an experimental protocol of emotion elicitation. We use a subset of the Continuously Anno-tated Signals of Emotions (CASE) dataset, including self-assessed ratings from thirty healthy subjects while watching two video clips: one fear-inducing and one relaxing. We analyse intrinsic irregularity and complexity of arousal time-series, performing Sample Entropy and Distribution Entropy algorithms. Results show a significantly higher complexity of time-varying emotion perception during the scary video compared to the relaxing video. Our findings, although preliminary, highlight a promising field of application of chaos theory methodologies to continuous emotion ratings, which can be exploited for the prediction of pathological moods in ecological settings.