A novel STA-EEGNet combined with channel selection for classification of EEG evoked in 2D and 3D virtual reality

Area of Science:

• Neuroscience
• Computer Science
• Human-Computer Interaction

Background:

• Virtual reality (VR) offers immersive experiences, enhancing engagement in gaming, education, and healthcare.
• Understanding neural responses to 2D vs. 3D VR stimuli is crucial for advancing brain-computer interfaces (BCIs).
• Existing EEG classification methods require enhancement for complex VR environments.

Purpose of the Study:

• To investigate electroencephalography (EEG) classification in response to 2D and 3D VR stimuli.
• To develop an advanced model for improved EEG signal classification within VR.
• To explore the impact of spatial-temporal attention and channel selection on VR-evoked EEG data.

Main Methods:

• Introduction of STA-EEGNet, a novel model enhancing EEGNet with spatial-temporal attention (STA).
• Utilization of one-way analysis of variance (ANOVA) for optimizing channel selection.
• Comparative experiments to evaluate STA-EEGNet performance against traditional EEGNet.

Main Results:

• STA-EEGNet demonstrated superior performance in classifying VR-evoked EEG signals compared to the standard EEGNet.
• A peak accuracy of 99.78% was achieved using STA-EEGNet with optimized channel selection.
• Spatial-temporal attention and optimal channel selection significantly benefit EEG classification in VR.

Conclusions:

• Integrating spatial-temporal attention with compact convolutional neural networks like EEGNet enhances EEG signal classification.
• The findings support the advancement of neural decoding and optimization of BCI applications.
• STA-EEGNet offers a promising approach for interpreting neural data from immersive VR environments.