A Data Augmentation Method for Motor Imagery EEG Signals Based on DCGAN-GP Network

Abstract

Motor imagery electroencephalography (EEG) signals have garnered attention in brain-computer interface (BCI) research due to their potential in promoting motor rehabilitation and control. However, the limited availability of labeled data poses challenges for training robust classifiers. In this study, we propose a novel data augmentation method utilizing an improved Deep Convolutional Generative Adversarial Network with Gradient Penalty (DCGAN-GP) to address this issue. We transformed raw EEG signals into two-dimensional time-frequency maps and employed a DCGAN-GP network to generate synthetic time-frequency representations resembling real data. Validation experiments were conducted on the BCI IV 2b dataset, comparing the performance of classifiers trained with augmented and unaugmented data. Results demonstrated that classifiers trained with synthetic data exhibit enhanced robustness across multiple subjects and achieve higher classification accuracy. Our findings highlight the effectiveness of utilizing a DCGAN-GP-generated synthetic EEG data to improve classifier performance in distinguishing different motor imagery tasks. Thus, the proposed data augmentation method based on a DCGAN-GP offers a promising avenue for enhancing BCI system performance, overcoming data scarcity challenges, and bolstering classifier robustness, thereby providing substantial support for the broader adoption of BCI technology in real-world applications.

Keywords: Generative Adversarial Networks; data augmentation; motor imagery electroencephalography signals; time–frequency maps.

Figure 1

Illustrates the proposed data augmentation method. It involves using the Short-Time Fourier Transform (STFT) to obtain time-frequency images of input EEG signals. Real data is used to train the Deep Convolutional Generative Adversarial Network-Gradient Penalty (DCGAN-GP) model to generate synthetic time-frequency images. These synthetic images are then mixed with real images in proportion and used to train a convolutional classifier to distinguish between left-hand and right-hand motor imagery (MI) actions.

Figure 2

DCGAN-GP model. The green color in the figure represents the two components of the model, namely the generator and the discriminator. The orange color represents the data, and the arrows indicate the direction of data flow.

Figure 3

Illustration of the architecture of the generator network.

Figure 4

Illustration of the architecture of the discriminator network.

Figure 5

Example of the experiment process without feedback.

Figure 6

Example of the experiment process with feedback.

Figure 7

(a) Real sample of right-hand movement; (b) generated sample of right-hand movement; (c) real sample of left-hand movement; (d) generated sample of left-hand movement.

Figure 8

Classification performance with different proportions of generated samples.