A Filter Bank Typical Correlation Analysis Method Based on Task-Related Components for SSVEP-BCI

Brain computer interface (BCI) systems have received widespread attention in recent years. Among them, the steady-state visual evoked potential (SSVEP)-BCI system has become the most widely studied and applied system due to its lack of user training, high signal-to-noise ratio, and high information transmission rate. The existing SSVEP signal processing methods have achieved certain results, but their lack of robustness to electroencephalogram (EEG) signals containing noise and artifacts, as well as insufficient utilization of harmonic information, make it difficult to improve their motion intention classification performance. This paper proposes a filter bank typical correlation analysis method based on task-related components (TC-FBCA) for designing SSVEP-BCI. The method extracts important EEG components by constructing task-related spatial filters, thereby improving the signal-to-noise ratio of EEG signals. Subsequently, the filtered EEG signal is decomposed into multiple sub bands, and typical correlation analysis is performed in each sub band to obtain the correlation coefficient of each band. Finally, we weighted the correlation coefficients of different frequency bands and selected the response frequency with the highest correlation coefficient as the subject’s exercise intention. We design a steady-state visual stimulation EEG signal acquisition experiment to validate the performance of the proposed method. The experimental results show that the average classification accuracy of TC-FBCA on single-subject reaches $86.48 \%$, achieving the most advanced classification accuracy. In addition, we combine the SSVEP-BCI system with our lower limb exoskeleton robot to build a real-time control platform, and conduct online testing of TC-FBCA. The average accuracy reaches $\mathbf{8 2. 1 0 \%}$, further demonstrating the effectiveness and practicality of our proposed method.