A Binocular Vision SSVEP Brain-Computer Interface Paradigm for Dual-Frequency Modulation.

Yike Sun,Liyan Liang,Jingnan Sun,Xiaogang Chen,Runfa Tian,Yuanfang Chen,Lijian Zhang,Xiaorong Gao

DOI: https://doi.org/10.1109/tbme.2022.3212192

2023-01-01

Abstract:Objective: This study presents a novel braincomputer interface paradigm of dual-frequency modulated steady-state visual evoked potential (SSVEP), aiming to suppress the unpredictable intermodulation components in current applications. This paradigm is especially suitable for training-free scenarios. Approach: This study built a dual-frequency binocular vision SSVEP brain-computer interface system using circularly polarized light technology. Two experiments, including a 6-target offline experiment and a 40-target online experiment, were taken with this system. Meanwhile, an improved algorithm filter bank dual-frequency canonical correlation analysis (FBDCCA) was presented for the dual-frequency SSVEP paradigm. Main results: Energy analysis was conducted for 9 subjects in the 6-target dual-frequency offline experiment, among which the signal-to-noise ratio of target frequency components have increased by 2 dB compared to the one of unpredictable intermodulation components. Subsequently, the online experiment with 40 targets was conducted with 12 subjects. With this new dual-frequency paradigm, the online training-free experiment's average information transmission rate (ITR) reached 104.56 +/- 15.74 bits/min, which was almost twice as fast as the current best dual-frequency paradigm. And the average information transfer rate for offline training analysis of this new paradigm was 180.87 +/- 17.88 bits/min. Significance: These results demonstrate that this new dual-frequency SSVEP brain-computer interface paradigm can suppress the unpredictable intermodulation harmonics and generate higher quality responses while completing dual-frequency encoding. Moreover, its performance shows high ITR in applications both with and without training. It is thus believed that this paradigm is competent for achieving large target numbers in braincomputer interface systems and has more possible practices.