Lightweight and Efficient Motor Imagery Decoding for High-ITR Multi-task Brain-Computer Interface

Brain-Computer Interface (BCI), a technology that enables the direct communication and control between the brain and external devices, have gained significant attention for their potential in various applications, many of which require real-time or online systems. While sufficient hardware performance is essential, optimizing the system's responsiveness largely hinges on minimizing the time window of input data used for analysis. In this paper, we propose a novel method, Time Windows and Frequency bands-Based Convolutional Neural Network (TWFB-CNN), designed to classify motor imagery tasks with a high degree of accuracy using only a 1-second time window. To validate the effectiveness of the proposed method, we conducted experiments using the Neuroscan system to collect electroencephalographic (EEG) data from 20 healthy subjects, each performing seven distinct motor imagery tasks. This dataset, combined with the TWFB-CNN algorithm, is aimed at advancing the state of brain-computer interfaces (BCIs) by enhancing both the classification accuracy and the feasibility of implementing real-time BCIs in practical applications. The results and framework presented here provide a robust foundation for future developments in online BCI systems, particularly in contexts that demand both rapid processing and high precision.