Biosignal Data Augmentation Based on Generative Adversarial Networks.

Area of Science:

• Artificial Intelligence
• Biomedical Engineering
• Data Science

Background:

• Generative Adversarial Networks (GANs) offer a powerful framework for learning generative models.
• Time-series data, such as biosignals, present unique challenges for traditional data generation methods.
• Data augmentation is crucial for enhancing the performance of machine learning models in classification tasks.

Purpose of the Study:

• To propose a novel synthetic data generation method for time-series biosignals.
• To adapt the Generative Adversarial Network (GAN) framework for time-series data.
• To evaluate the effectiveness of the proposed method for data augmentation in biosignal classification.

Main Methods:

• Developed a synthetic data generation method utilizing Generative Adversarial Networks (GANs).
• Implemented recurrent neural networks with long short-term memories for both generator and discriminator networks within the GAN framework.
• Applied the GAN-based method to generate synthetic biosignals for augmentation purposes.

Main Results:

• Successfully generated realistic synthetic biosignals using electrocardiogram (ECG) and electroencephalogram (EEG) datasets.
• Demonstrated the capability of the proposed GAN-based method to produce high-fidelity time-series data.
• Confirmed the effectiveness of the synthetic data generated by the method for data augmentation in biosignal classification.

Conclusions:

• The proposed GAN-based method is effective for generating synthetic time-series biosignals.
• The method successfully adapts GANs for time-series data generation using recurrent neural networks.
• Data augmentation with GAN-generated biosignals significantly improves biosignal classification performance.