SAST-GCN: Segmentation Adaptive Spatial Temporal-Graph Convolutional Network for P3-Based Video Target Detection.

Area of Science:

• Neuroscience
• Computer Science
• Signal Processing

Background:

• Detecting video-induced P3 signals is essential for brain-computer interface (BCI) based video target detection systems.
• Brain response patterns for video-induced P3 are dynamic and involve interactions across multiple brain regions.

Purpose of the Study:

• To propose a novel Segmentation Adaptive Spatial-Temporal Graph Convolutional Network (SAST-GCN) for enhanced P3-based video target detection.
• To leverage the dynamic characteristics of P3 signal data by segmenting it according to processing stages and constructing corresponding brain network connections.

Main Methods:

• The proposed SAST-GCN utilizes adaptive spatial-temporal graph convolution to extract spatio-temporal features from EEG data.
• Data segmentation based on P3 processing stages allows for dynamic brain network construction.
• A style-based recalibration module is incorporated to enhance feature extraction by selecting high-contribution feature maps.

Main Results:

• Experimental results show the superiority of the SAST-GCN model compared to existing baseline methods.
• Ablation experiments confirm that data segmentation for brain connection construction significantly improves recognition performance.
• The model accurately reflects dynamic connection relationships between EEG channels.

Conclusions:

• The SAST-GCN model offers a significant advancement in P3-based video target detection for BCI systems.
• Dynamic modeling of brain network connections through data segmentation is crucial for improving EEG signal recognition.
• The proposed method demonstrates high potential for real-world BCI applications requiring accurate video target identification.