Music-induced Emotion Oscillation Modeling

Music can evoke emotional responses, but inter pretable models of music-induced emotional oscillations remain limited. We propose a reproducible framework based on frontal alpha asymmetry (FAA). Two FAA-derived dynamic features, the Negative Emotional Index (NEI) and the Fluctuating Emotional Index (FEI), summarize the signed temporal mean and temporal variation of nonlinearly mapped FAA. We evaluated their group-related and valence-related information through two exploratory validation tasks in 28 patients with major depressive disorder and 30 healthy controls. NEI and FEI differed between groups; within the patient group, both were lower under positive than negative valence labels. Combining static spectral-energy distribution features with FAA-derived dynamic features yielded depression-recognition accuracy of 87.5% and class-balanced mean accuracy for music-induced emotion recognition of 80.9% on participant-disjoint held-out sets. These findings provide preliminary within-dataset evidence that static and dynamic features carry complementary information for depression and emotion recognition and may help characterize music-induced emotional states and their temporal stability. Larger external cohorts and prospective studies are needed to assess generalizability and clinical applicability.