Feature Selection via Dynamic Graph-based Attention Block in MI-based EEG Signals

TL;DR

The proposed end-to-end deep preprocessing method effectively enhances MI characteristics while attenuating features with low correlation to MI characteristics, and it was demonstrated that the proposed method could enhance discriminative features related to MI characteristics.

Abstract

Brain-computer interface (BCI) technology enables direct interaction between humans and computers by analyzing brain signals. Electroencephalogram (EEG) is one of the non-invasive tools used in BCI systems, providing high temporal resolution for real-time applications. However, EEG signals are often affected by a low signal-to-noise ratio, physiological artifacts, and individual variability, representing challenges in extracting distinct features. Also, motor imagery (MI)-based EEG signals could contain features with low correlation to MI characteristics, which might cause the weights of the deep model to become biased towards those features. To address these problems, we proposed the end-to-end deep preprocessing method that effectively enhances MI characteristics while attenuating features with low correlation to MI characteristics. The proposed method consisted of the temporal, spatial, graph, and similarity blocks to preprocess MI-based EEG signals, aiming to extract more discriminative features and improve the robustness. We evaluated the proposed method using the public dataset 2a of BCI Competition IV to compare the performances when integrating the proposed method into the conventional models, including the DeepConvNet, the M-ShallowConvNet, and the EEGNet. The experimental results showed that the proposed method could achieve the improved performances and lead to more clustered feature distributions of MI tasks. Hence, we demonstrated that our proposed method could enhance discriminative features related to MI characteristics.

Figures (2)

• Figure 1: Visualization of the overall process in the proposed method. EEG signals of rest and MI are used as input simultaneously. The temporal, graph, and spatial blocks extract the significant features to effectively calculate the negative similarity. The similarity block contains a processing step that generates the value $\textit{v}_i$ to enhance MI features while attenuating rest features. ($D$: the dimension of the feature vector, $R$: the number of time steps compressed from rest signals, and $M$: the number of time steps compressed from MI signals).
• Figure 2: Feature visualization using the t--SNE for A5 on BCIC2a. The colors represented the types of each MI task. (w/o: without the proposed method and w/: with the proposed method).