Classification of Motor Imagery Data Using TCN-Transformer

Brain-Computer Interfaces (BCIs) represent a rapidly advancing field that enables communication between the brain and external devices. Motor Imagery (MI) and Motor Execution (ME) paradigms are particularly valuable for BCI applications, especially in neurorehabilitation contexts. This study introduces a novel hybrid TCN-Transformer architecture designed to automatically classify electroencephalography (EEG) signals from MI and ME tasks. The proposed model combines a Temporal Convolutional Network (TCN) for feature extraction with a Transformer encoder for modeling global temporal dependencies. We validated our approach through comprehensive experiments on three datasets: Shuqfa-103 (103 subjects), Brunner-9 (9 subjects), and Kodera-29 (29 subjects). The model demonstrated strong performance across multiple classification scenarios, achieving accuracies of 84.84% for 2-class classification and 64.57% for 4-class classification on the Shuqfa-103 dataset. Comparative analysis with existing studies shows the promising performance of the proposed model. Furthermore, the comparison of the classification results on separated MI and ME datasets was performed. The model was successfully integrated into a real-world neuroinformatics laboratory workflow, demonstrating its practical applicability for real-time EEG signal processing and classification.