Real-time channel selection for enhanced steady-state visual evoked potentials online brain-computer interface systems.

BACKGROUND

In recent years, researchers have actively developed new brain-computer interface (BCI) systems while seeking optimal channel combinations. Although channel selection has advanced significantly in BCI systems, few studies have specifically addressed channel selection for steady-state visual evoked potentials-BCI (SSVEP-BCI).

NEW METHOD: This study proposes an online SSVEP-BCI system with dynamic channel selection during experiments. The proposed method constructs a multidimensional feature framework that incorporates signal energy, stability, and inter-channel correlation. The abnormality of each feature is then quantified to generate corresponding anomaly scores. These anomaly scores are integrated through a hierarchical decision mechanism to produce a comprehensive channel quality score. Based on this score, bad channels are precisely identified and removed, enabling training-free and dynamic channel selection.

RESULTS

The proposed multi-Adaptive priority-based SSVEP channel selection (MAPS-CS) method achieves the best performance compared with three existing channel selection methods. For the standard filter bank canonical correlation analysis (FBCCA), the accuracy of the proposed method was increased for four stimulus durations. Compared with the channel ensemble (CE) method, the proposed method achieved accuracy improvements of 3.5%, 4.1%, 4.4%, and 6.5% for stimulus durations of 2 s, 1.5 s, 1 s, and 0.5 s, respectively.

COMPARISON WITH EXISTING

METHODS

The proposed method provides the best performance in FBCCA compared with CE, binary harmony search (BHS) and TOP-K local optimization channel selection (TOP-K-LOCS).

CONCLUSIONS

These results confirm that the system can effectively detect bad channels in laboratory-based online experiments.