Abstract:Brain-Computer Interfaces (BCIs) rely on accurately decoding electroencephalography (EEG) motor imagery (MI) signals for effective device control. Graph Neural Networks (GNNs) outperform Convolutional Neural Networks (CNNs) in this regard, by leveraging the spatial relationships between EEG electrodes through adjacency matrices. The EEG_GLT-Net framework, featuring the state-of-the-art EEG_GLT adjacency matrix method, has notably enhanced EEG MI signal classification, evidenced by an average accuracy of 83.95% across 20 subjects on the PhysioNet dataset. This significantly exceeds the 76.10% accuracy rate achieved using the Pearson Correlation Coefficient (PCC) method within the same framework.

What problem does this paper attempt to address?

The problem this paper attempts to address is improving the accuracy of Electroencephalogram (EEG) Motor Imagery (MI) signal classification by combining Graph Neural Networks (GNN) and Reinforcement Learning (RL). Specifically, the paper proposes a new algorithm—EEG RL-Net, aimed at overcoming the issue of classification inaccuracy at certain time points due to signal ambiguity in existing methods. ### Problem Background - **Brain-Computer Interface (BCI)**: BCI systems rely on accurately decoding EEG MI signals to achieve effective device control. - **Existing Methods**: Current methods such as Convolutional Neural Networks (CNN) and GNN based on Pearson Correlation Coefficient (PCC) perform well in EEG MI signal classification but still have room for improvement. - **EEG GLT-Net**: One of the state-of-the-art methods, which significantly improves classification accuracy by optimizing the adjacency matrix density, but still faces issues with signal ambiguity at certain time points. ### Research Objectives - **Improve Classification Accuracy**: Further enhance the accuracy of EEG MI signal classification by introducing RL methods. - **Handle Ambiguous Signals**: Effectively identify and classify time points where the signals are more ambiguous. ### Methodological Innovations - **EEG RL-Net**: Combines pre-trained GNN blocks and RL blocks, where the GNN block is used to extract graph features, and the RL block is used to decide whether to classify or skip the current time point. - **Dueling DQN**: Utilizes the Dueling DQN network to optimize the decision-making process of the RL agent, improving classification speed and accuracy. ### Main Contributions 1. **EEG RL-Net**: Proposes a new method where the RL agent decides whether to classify each time point, thus enabling classification as quickly as possible within a predefined episode length. 2. **Optimal Reward and Maximum Episode Length Settings**: Evaluates the accuracy and classification speed under different reward settings and maximum episode lengths, determining the best combination. 3. **Performance Validation**: Conducts performance evaluation on the PhysioNet dataset with 20 subjects, showing significant improvements in both accuracy and efficiency with EEG RL-Net. ### Conclusion By introducing RL methods, this paper successfully addresses the issue of classification inaccuracy due to signal ambiguity at certain time points in existing EEG MI signal classification methods, significantly improving both accuracy and efficiency. This innovation provides new ideas and methods for the development of brain-computer interface technology.

EEG_RL-Net: Enhancing EEG MI Classification through Reinforcement Learning-Optimised Graph Neural Networks

EEG_GLT-Net: Optimising EEG Graphs for Real-time Motor Imagery Signals Classification

An AMCMD Approach for Robust EEG Signal Classification: Unraveling Neurophysiological Insights and BCI Applications

GCNs-Net: A Graph Convolutional Neural Network Approach for Decoding Time-resolved EEG Motor Imagery Signals

An Extended Computer Aided Diagnosis System for Robust BCI Applications

MBGA-Net: A multi-branch graph adaptive network for individualized motor imagery EEG classification

RLEEGNet: Integrating Brain-Computer Interfaces with Adaptive AI for Intuitive Responsiveness and High-Accuracy Motor Imagery Classification

Advanced TSGL-EEGNet for Motor Imagery EEG-Based Brain-Computer Interfaces

EEG-BBNet: a Hybrid Framework for Brain Biometric using Graph Connectivity

EEG-DCNet: A Fast and Accurate MI-EEG Dilated CNN Classification Method

Classification Of Eeg Signal Using Convolutional Neural Networks

Kernel-Based Regularized EEGNet Using Centered Alignment and Gaussian Connectivity for Motor Imagery Discrimination

Optimizing EEG Signal Classification in BCI Systems: A Comparative Study of Hybrid Transfer Learning and Advanced Machine Learning Approaches

IFBCLNet: Spatio-temporal frequency feature extraction-based MI-EEG classification convolutional network

EEG-GNN: Graph Neural Networks for Classification of Electroencephalogram (EEG) Signals

Enhancing EEG Motor Imagery Decoding Performance Via Deep Temporal-domain Information Extraction

Deep Fusion Feature Learning Network for MI-EEG Classification.

Enhancing Motor Imagery EEG Signal Classification with Simplified GoogLeNet.

Graph neural network based on brain inspired forward-forward mechanism for motor imagery classification in brain-computer interfaces

An in-depth survey on Deep Learning-based Motor Imagery Electroencephalogram (EEG) classification

A Spiking Neural Network With Adaptive Graph Convolution and LSTM for EEG-Based Brain-Computer Interfaces