Multi-Domain Feature Fusion and Channel Attention in an Inception-Based Architecture for Motor Imagery EEG Decoding
TL;DR
The results reveal that the proposed framework can distinguish motor imagery patterns across different limb regions and effectively classify predictive EEG features between symmetric limbs performing identical imagined movements, providing an offline proof-of-concept for multi-limb BCI control.
Abstract
Highlights What are the main findings? Evaluated under a rigorous intra-subject (subject-dependent) cross-validation protocol across 292 individuals, the SE-EEG-Inception model achieved a mean 4-class accuracy of 89.4%. Binary classification accuracies reached 88.3% for left vs. right arm and 90.0% for left vs. right leg. These results reveal that the proposed framework can distinguish motor imagery patterns across different limb regions and effectively classify predictive EEG features between symmetric limbs performing identical imagined movements. What are the implications of the main findings? Non-invasive EEG can be used to identify distinct motor imagery patterns and classify predictive signals between symmetric limbs (e.g., left vs. right arm, left vs. right leg), laying the foundation for developing multi-limb motor imagery brain–computer interface systems. These findings may support future applications in intelligent prosthetic control and neurorehabilitation. Abstract Background/Objectives: Existing motor imagery (MI) EEG decoding is often limited by small datasets, affecting generalization reliability. This study aims to robustly decode multi-limb MI intentions. Methods: We collected an MI-EEG dataset from 292 participants (242 young adults, 50 older adults) performing left/right-arm and left/right-leg imagery. After extracting time-, frequency-, and channel correlation features, we proposed an SE-EEG-Inception model for classification. Results: Evaluated under a strict intra-subject cross-validation protocol, the model achieved a mean 4-class accuracy of 89.4%. For binary tasks, accuracies reached 88.3% (left vs. right arm) and 90.0% (left vs. right leg). Conclusions: The model successfully distinguishes predictive EEG features across different and symmetric limbs. Crucially, this high classification performance demonstrates data-driven predictive utility rather than mechanistic proof of neural differences, providing an offline proof-of-concept for multi-limb BCI control.