Physiological Correlates of Emotional States: A Multimodal Signal Analysis

Emotional responses involve coordinated changes across multiple components of the autonomic nervous system. However, studies simultaneously combining cardiovascular, electrodermal, and respiratory measures remain limited, particularly in video-clip paradigms, despite their ecological validity for emotion elicitation. This thesis provides a multimodal characterization of the physiological responses elicited by video clips representing distinct affective states. The study involved the simultaneous acquisition of cardiovascular, electrodermal, and respiratory signals from ten healthy volunteers. Participants were exposed to video clips belonging to six affective categories, sampled from the E-MOVIE database (Maffei & Angrilli, 2019): Fear, Sadness, Compassion, Erotic, Scenery, and Neutral. After each clip, participants rated the emotional valence and arousal. Signal processing enabled the extraction of 11 physiological features (3 cardiovascular, 3 electrodermal, 5 respiratory). Responses to each non-neutral category were compared with those elicited by Neutral clips and with patterns reported in the literature. Autonomic differentiation depended on both emotional category and physiological system. Cardiovascular features were the most informative, with heart rate differing significantly from Neutral in most categories. Respiratory features were particularly informative for the Erotic category, while electrodermal activity proved to be the least sensitive modality. Subjective valence and arousal ratings were strongly correlated with E-MOVIE ratings, supporting the consistency of the emotional responses. Overall, the findings support the value of combining cardiovascular, electrodermal, and respiratory measures to obtain a more nuanced characterization of emotion-related autonomic responses. These results also provide a basis for future machine-learning models for emotion recognition.