Emotion Recognition Based on EEG Using Generative Adversarial Nets and Convolutional Neural Network.

Area of Science:

• Neuroscience
• Computer Science
• Artificial Intelligence

Background:

• Emotion recognition is crucial for human-computer interaction (HCI).
• Automatic emotion recognition using electroencephalography (EEG) is key for brain-computer interface (BCI) applications.
• Deep learning models show promise but struggle with limited and imbalanced EEG datasets.

Purpose of the Study:

• To address the challenges of small sample sizes and category imbalance in EEG datasets for emotion recognition.
• To introduce a novel sample generation method using generative adversarial networks (GANs) for data augmentation.
• To evaluate the impact of GAN-based data augmentation on deep learning models for EEG emotion recognition.

Main Methods:

• Proposed a generative adversarial network (GAN) based approach for augmenting EEG data.
• Compared the performance of frequency band correlation and frequency band separation computational models.
• Evaluated models on standard EEG-based emotion recognition datasets before and after data augmentation.

Main Results:

• GAN-based data augmentation significantly improved the performance of deep learning models for EEG emotion recognition.
• The frequency band correlation deep learning model demonstrated superior performance in emotion recognition tasks.
• Data augmentation effectively mitigated issues related to small sample size and class imbalance in EEG datasets.

Conclusions:

• Generative adversarial networks are effective for augmenting EEG data, enhancing emotion recognition performance.
• Frequency band correlation models are more suitable for EEG-based emotion recognition compared to frequency band separation.
• The proposed method offers a viable solution for improving deep learning applications in BCI and HCI.