Deep Learning Classification of Neuro-Emotional Phase Domain Complexity Levels Induced by Affective Video Film Clips

Area of Science:

• Neuroscience
• Affective Computing
• Signal Processing

Background:

• Classifying discrete emotions from electroencephalography (EEG) signals is challenging.
• Existing methods often rely on regional neuro-cortical connectivity, which may not fully capture emotional nuances.
• Developing novel markers for accurate emotion recognition is crucial for understanding affective states.

Purpose of the Study:

• To propose and validate a novel emotional complexity marker for classifying discrete emotions induced by video stimuli.
• To compare the efficacy of local neuronal complexity estimation against regional neuro-cortical connectivity (Phase Locking Value - PLV) in emotion clustering.
• To investigate the influence of gender and individual experiences on emotion processing using EEG.

Main Methods:

• Principal Component Analysis (PCA) was applied to phase space trajectory matrices (PSTM) from 6-second EEG segments to derive local neuronal complexity.
• Phase Locking Value (PLV) was calculated to assess interhemispheric connectivity.
• Long-Short-Term Memory (LSTM) networks were employed for deep learning-based classification of nine discrete emotions.
• Data from healthy male and female participants (aged 22-33) viewing affective video clips were analyzed.

Main Results:

• The proposed emotional complexity marker achieved high classification accuracy ([Formula: see text]) for discriminating positive from negative emotions.
• Significant differences were observed between genders in amusement using both complexity markers and PLV (p << 0.5).
• Local neuronal complexity was primarily sensitive to affective valence, while PLV correlated more with affective arousal.

Conclusions:

• Local neuronal complexity is a superior marker for clustering discrete emotions compared to regional neuro-cortical connectivity.
• Emotion formation is predominantly influenced by individual experiences, not gender, across most emotional states.
• The developed complexity markers show promise for accurate emotion recognition and understanding affective neuroscience.