Leveraging Graph Neural Networks to Decode Music-Induced Emotions from EEG.

Area of Science:

• Neuroscience
• Machine Learning
• Affective Computing

Background:

• Decoding emotion processing from electroencephalogram (EEG) signals is complex.
• Understanding neural correlates of music perception and enjoyment is crucial for affective computing.

Purpose of the Study:

• Investigate Graph Neural Networks (GNNs) for interpreting EEG data.
• Classify emotional states (enjoyment vs. disenjoyment) during music listening using EEG.
• Compare GNN performance against traditional Convolutional Neural Networks (CNNs).

Main Methods:

• Utilized the Naturalistic Music EEG Dataset-Tempo (NMED-T).
• Applied GNNs to EEG recordings to identify neural activity patterns.
• Performed classification of emotional states based on music exposure.

Main Results:

• GNNs achieved 87% accuracy in classifying previously unseen EEG data.
• GNNs demonstrated superior performance compared to CNNs for this task.
• Successfully distinguished neural patterns associated with music enjoyment and disenjoyment.

Conclusions:

• GNNs show significant potential for decoding intricate brain signals related to emotion processing.
• GNNs offer a promising approach for analyzing EEG data in affective neuroscience.
• Future work may use GNNs to identify brain regions involved in music-induced emotions.