DualDomain-AttenNet: Synergizing time–frequency analysis and attention mechanisms for Motor Imagery BCI enhancement

This paper introduces DualDomain-AttenNet, a novel deep learning model for brain–computer interfaces (BCI) based on electroencephalography (EEG) data, focusing on integrating time and frequency domain analyses with attention mechanisms. The model addresses challenges in EEG signal modeling, particularly in Motor Imagery (MI) BCIs, by enhancing feature extraction and signal classification. We present a unique architecture that combines temporal convolution networks with global filters, enabling the simultaneous analysis of EEG signals in both the time and frequency domains. The model's incorporation of multi-head self-attention mechanisms allows it to adaptively focus on significant signal features, improving the reliability of EEG interpretation. The proposed model outperforms state-of-the-art models in MI classification, demonstrating robustness across various subject data and highlighting its potential for personalized BCI applications. The experimental results confirm the model's effectiveness through an extensive ablation study, emphasizing the capabilities of the model in complex EEG data processing. The findings indicate that DualDomain-AttenNet could significantly advance the field of BCI, offering a robust framework for EEG signal analysis and opening new avenues for applications in rehabilitation, neurogaming, and human–computer interaction.