Differentiating brain states via multi-clip random fragment strategy-based interactive bidirectional recurrent neural...

Area of Science:

• Neuroscience and Artificial Intelligence
• Brain-Computer Interface (BCI) research
• Signal processing for neural data

Background:

• Electroencephalography (EEG) is a non-invasive, low-cost method for brain study and BCI development.
• Differentiating brain states using EEG is crucial for understanding brain mechanisms.
• Current Recurrent Neural Network (RNN) models for EEG brain state classification show limitations in performance.

Purpose of the Study:

• To improve brain state classification accuracy using EEG signals.
• To address limitations in EEG signal organization and RNN architecture design.
• To introduce a novel model for enhanced brain state differentiation.

Main Methods:

• Proposed a novel Multi-Clip Random Fragment Strategy-based Interactive Bidirectional Recurrent Neural Network (McRFS-IBiRNN) model.
• Developed the McRFS component to reorganize input EEG signals for RNN suitability.
• Designed the IBiRNN component with interaction connections to fuse bidirectional features.

Main Results:

• Achieved high classification accuracies: 96.97% (individual) and 99.34% (group) for four-category classification on an EEG motor/imagery dataset.
• Demonstrated strong generalization with 99.01% accuracy on unseen subjects.
• Outperformed existing methods in brain state differentiation tasks.

Conclusions:

• The proposed McRFS-IBiRNN model significantly improves brain state classification performance.
• The novel signal organization and RNN architecture enhance EEG-based BCI capabilities.
• This model shows great superiority and generalization for differentiating brain states from EEG signals.