A Brain-Computer Interface Based on Multifocal SSVEPs Detected by Inter-Task-Related Component Analysis

Multifocal steady-state visual evoked potentials (mfSSVEPs) have been successfully applied to assess visual field loss in glaucoma. However, the potential of mfSSVEPs for command control has not been fully explored yet. It is significant to detect single-trial mfSSVEPs and establish a brain-computer interface (BCI) system. This study designed a stimulating paradigm that contains 32 targets, with each target composed of five fan-shaped flickers in a circle. The five flickers were modulated by five frequencies and formed a five-bit binary encoding system through controlling the ON/OFF state of each flicker. Twelve subjects participated in an offline and an online experiments. Inter-task-related component analysis (iTRCA) combined with a probabilistic model was proposed for target recognition. Notably, the training data needed for calibration corresponded to only six out of the 32 targets. It was found that the increasing number of flickers showed a negative impact on the mfSSVEP signal. The accuracy reached 80.9% ± 11.7% on average with a peak of 95.3% by iTRCA, which was significantly higher than that by a traditional method. The results indicate that the proposed stimulation and algorithm are effective for encoding and decoding BCI commands. Therefore, the mfSSVEP-based BCI enables the augmentation of the BCI instruction set without any burden of collecting extra training data.