EEG Data Augmentation for Emotion Recognition Using Diffusion Model.

Area of Science:

• Affective Computing
• Machine Learning
• Signal Processing

Background:

• Electroencephalogram (EEG) signals are crucial for emotion recognition in affective computing.
• Collecting high-quality EEG data is complex and resource-intensive.
• Traditional generative models face challenges in EEG signal generation.

Purpose of the Study:

• To apply diffusion models for generating high-quality EEG signals for emotion recognition.
• To optimize diffusion models for EEG signal enhancement.
• To investigate the impact of generated data quantity on emotion recognition performance.

Main Methods:

• Utilized diffusion models, optimizing the sampling stage for EEG signal generation.
• Applied data augmentation techniques to enhance original EEG signals.
• Evaluated the method on SEED, SEED-IV, and DEAP datasets.

Main Results:

• The proposed data augmentation method significantly improved emotion recognition accuracy.
• Diffusion models demonstrated superior performance in generating high-quality EEG signals compared to traditional methods.
• The quantity of generated data positively impacted task performance.

Conclusions:

• Diffusion models offer a promising approach for efficient and effective EEG data augmentation in emotion recognition.
• Optimized diffusion models can overcome limitations in EEG data collection for affective computing.
• This method provides a scalable solution for enhancing emotion recognition systems.