An optimized EEGNet decoder for decoding motor image of four class fingers flexion

Area of Science:

• Neuroscience
• Biomedical Engineering
• Machine Learning

Background:

• Brain-computer interfaces (BCI) offer potential for individuals with motor function loss.
• Accurate decoding of electroencephalogram (EEG) signals is crucial for non-invasive BCI applications.
• Sensory motor rhythm (SMR) analysis is key for decoding motor intentions.

Purpose of the Study:

• To enhance the accuracy of decoding EEG signals for index and thumb finger movements.
• To investigate the application of an explanatory convolutional neural network (EEGNet) with a SimAM attention module.
• To analyze event-related desynchronization (ERD) and event-related synchronization (ERS) patterns.

Main Methods:

• Utilized an EEGNet model incorporating the SimAM attention module.
• Collected EEG data from eight healthy subjects during index and thumb finger movements (left and right hands).
• Analyzed data from one second before to two seconds after the motor action.

Main Results:

• Achieved an average classification accuracy of 72.91% in decoding finger movements.
• Successfully decoded EEG signals related to index and thumb movements.
• Characterized ERD and ERS patterns associated with finger movements.

Conclusions:

• The proposed EEGNet with SimAM attention significantly improves EEG signal decoding accuracy for BCI.
• Findings support the potential for fine-grained control of external devices and rehabilitation equipment.
• This research advances non-invasive BCI technology for assistive applications.