BACNet: A multi-attention network for cross-subject and cross-task EEG-based pilot operational intent recognition.

Area of Science:

• Neuroscience
• Aerospace Engineering
• Computer Science

Background:

• Pilot operational intent recognition is vital for flight safety and human-machine interaction.
• Electroencephalography (EEG) offers high temporal resolution for non-invasive intent recognition.
• Existing EEG methods face challenges with model complexity and feature extraction accuracy.

Purpose of the Study:

• To develop an efficient and accurate neural network for pilot intent recognition using EEG signals.
• To address limitations in current EEG-based intent recognition approaches.

Main Methods:

• Introduction of the Balanced Attention Convolutional Network (BACNet) for enhanced EEG-based intent recognition.
• BACNet utilizes a three-branch parallel convolutional structure for multi-scale time-frequency feature extraction.
• Dynamic feature modulation adaptively highlights salient channels and spatial locations.

Main Results:

• BACNet achieved 96.07% average classification accuracy in a three-class intent recognition task.
• The model significantly outperformed five state-of-the-art baseline methods.
• Ablation studies confirmed the effectiveness of BACNet's collaborative attention design.

Conclusions:

• BACNet offers a lightweight and highly accurate solution for pilot intent recognition.
• The proposed architecture demonstrates broad applicability in brain-computer interface (BCI) systems.