PSAEEGNet: pyramid squeeze attention mechanism-based CNN for single-trial EEG classification in RSVP task

Area of Science:

• Neuroscience
• Machine Learning
• Biomedical Engineering

Background:

• Accurate single-trial electroencephalogram (EEG) classification is vital for EEG-based target recognition in rapid serial visual presentation (RSVP) tasks.
• P300 signals are key for RSVP tasks, but suffer from low signal-to-noise ratio and limited data.

Purpose of the Study:

• To optimize convolutional neural networks (CNNs) for improved P300 classification in single-trial EEG.
• To introduce PSAEEGNet, a novel CNN model designed for enhanced feature extraction in EEG data.

Main Methods:

• Developed PSAEEGNet, integrating standard convolutional layers, pyramid squeeze attention (PSA) modules, and deep convolutional layers.
• Focused on extracting temporal and spatial features of P300 signals to a finer granularity.

Main Results:

• PSAEEGNet demonstrated significantly improved performance compared to existing single-trial EEG classification methods.
• Achieved a mean true positive rate of 0.7949 and a mean Area Under the Curve (AUC) of 0.9341 (p < 0.05).

Conclusions:

• The proposed model effectively extracts temporal and spatial features of P300, leading to more accurate single-trial EEG classification.
• PSAEEGNet shows potential for enhancing EEG-based target recognition systems and advancing practical applications.