Fractal Spiking Neural Network Scheme for EEG-Based Emotion Recognition

Area of Science:

• Neuroscience
• Artificial Intelligence
• Biomedical Engineering

Background:

• Electroencephalogram (EEG)-based emotion recognition is crucial for clinical applications.
• Existing methods often fail to fully leverage temporal, spectral, and spatial EEG data.
• There is a need for advanced techniques to improve the accuracy and comprehensiveness of EEG emotion recognition.

Purpose of the Study:

• To propose a novel Fractal Spike Neural Network (Fractal-SNN) for improved EEG-based emotion recognition.
• To exploit multi-scale Temporal-Spectral-Spatial (TSS) information inherent in EEG signals.
• To enhance the generalization capability of the proposed model.

Main Methods:

• Developed a Fractal-SNN block simulating biological neural structures using spiking neurons and a fractal rule.
• Implemented a novel training technique called inverted drop-path to improve model generalization.
• Extracted discriminative multi-scale TSS features from EEG signals.

Main Results:

• The proposed Fractal-SNN scheme demonstrated superior performance compared to advanced methods.
• Experiments were conducted on four public benchmark EEG databases (DREAMER, DEAP, SEED-IV, MPED).
• Subject-dependent protocols confirmed the effectiveness of the proposed approach.

Conclusions:

• The Fractal-SNN scheme offers a promising and effective solution for EEG-based emotion recognition.
• The method successfully integrates multi-scale TSS information for more accurate emotion classification.
• This approach has significant potential for clinical diagnosis, treatment, and rehabilitation.