Temporal Focal Modulation Networks for EEG-Based Cross-Subject Motor Imagery Classification

Abstract

Motor Imagery (MI) EEG decoding is crucial in Brain-Computer Interface (BCI) technology, facilitating direct communication between the brain and external devices. However, accurately capturing temporal dependencies in MI EEG signals, especially in subject-independent MI-BCIs, remains a persistent challenge. In this paper, we present Temporal-FocalNets, a novel framework designed to address this challenge by leveraging focal modulation techniques. Temporal-FocalNets efficiently prioritize temporal dynamics, thereby enhancing the accuracy and robustness of MI EEG decoding models. Through comprehensive experiments on benchmark datasets (2a and 2b), Temporal-FocalNets demonstrates superior performance compared to established baseline models. This innovation marks a significant advancement in subject-independent MI-BCIs, offering new possibilities for individuals with motor disabilities to interact with their environment using brain signals.