Abstract
INTRODUCTION: High frequency electroencephalogram (EEG) activity, particularly in the high gamma range, plays an important role in research on human emotions. However, the current understanding of high gamma EEG responses to emotional stimuli in virtual reality (VR) remains limited, especially regarding local activations and distributed network characteristics during different emotional states. METHODS: In this study, EEG responses to positive and negative VR stimuli were analyzed. EEG data were recorded from 19 participants as they viewed 4-second VR videos designed to elicit positive and negative responses. Two neural signatures were examined: high gamma band (53-80 Hz) spectral power and brain network features (nodal/local efficiency). RESULTS AND DISCUSSION: Spectral power analysis revealed valence-specific spatial patterns in spectral power, with significantly higher frontal gamma activity during positive states and increased right temporal gamma power during negative states. Network analysis revealed elevated local efficiency during positive emotions, indicating enhanced modular connectivity. Machine learning classification demonstrated higher accuracy for spectral power features (73.57% ± 2.30%) compared to nodal efficiency (69.51% ± 2.62%) and local efficiency (65.03% ± 1.33%), with key discriminators identified in frontal, temporal, and occipital regions. These findings suggest that localized high gamma activity provides more direct discriminative information for emotion recognition in VR than network topology metrics, advancing the understanding of neurophysiological responses in immersive VR environments.