Heterogeneous Recurrence Network Analysis: Unveiling Complex Dynamics for Emotion Recognition in EEG Signals

This study presents a novel approach for emotion recognition through the analysis of EEG signals, employing Heterogeneous Recurrence Network Analysis (HRNA). Recognizing the complex and dynamic nature of brain activities, HRNA is adept at capturing these intricacies using a sophisticated network topology that integrates both heterogeneous and homogeneous properties. This integration enhances the potential for machine learning applications in deciphering the brain's complex dynamics. We deployed HRNA to the SITU Emotion EEG dataset (SEED-IV), aiming to uncover the multifaceted temporal patterns associated with emotional states. Our methodology involved constructing advanced network features that reflect the heterogeneity of brain signals, thereby providing a more nuanced understanding of emotional responses. Our HRNA-based approach achieved an average AUG of 0.73, demonstrating its effectiveness in distinguishing diverse emotional features within EEG data. These results indicate HRNA's potential as a powerful tool for accurately measuring and categorizing emotional states through EEG. The implications of our findings extend to providing objective support for the clinical diagnosis and treatment of mental health disorders, highlighting HRNA's relevance in both research and clinical settings.