Decoding emotion with phase-amplitude fusion features of EEG functional connectivity network

Area of Science:

• Neuroscience
• Cognitive Science
• Biomedical Engineering

Background:

• Understanding emotional cognition is crucial for human-computer interaction.
• Existing electroencephalographic (EEG) methods for emotion recognition primarily use phase-based functional connectivity, neglecting amplitude information.
• This limits the capture of dynamic cortical oscillation energy fluctuations.

Purpose of the Study:

• To investigate the efficacy of amplitude-based functional networks (AEC) for representing emotional states.
• To develop an efficient phase-amplitude fusion framework (PAF) for enhanced emotion recognition.
• To extract and analyze fused spatial topological features for multi-class emotion classification.

Main Methods:

• Examined amplitude envelope correlation (AEC) for emotional state representation.
• Proposed a phase-amplitude fusion framework (PAF) integrating phase locking value (PLV) and AEC.
• Utilized common spatial pattern (CSP) for extracting fused spatial topological features.
• Conducted experiments on the DEAP and MAHNOB-HCI datasets.

Main Results:

• Amplitude-based functional networks (AEC) effectively characterize emotional states.
• Differential network patterns in AEC reflect dynamic interactions in emotion-related brain regions.
• The proposed phase-amplitude fusion (PAF) features significantly outperformed state-of-the-art methods in classification accuracy on both datasets.
• Learned spatial filters from PAF are interpretable, describing affective patterns from both phase and amplitude.

Conclusions:

• Amplitude-based connectivity provides valuable insights into emotional states.
• The novel phase-amplitude fusion framework (PAF) offers a more comprehensive approach to EEG-based emotion recognition.
• This fusion method enhances classification accuracy and provides interpretable affective activation patterns.