Detection Performance Analysis based on EEG Signal for Visual BCI

The brain computer interfaces (BCIs) offer a possibility of communication for people with computer. Event related potential (ERP) and steady-state visual evoked potential (SSVEP) studies using simple images were mainly conducted to observe brain response characteristics to visual stimuli. In this paper, we present a Visual BCI dataset according to the presence or absence of a target in a still image and analyze the performance of various feature extraction and classification algorithms. Throughout various experiments, the proposed variance and mean ratio (VMR) based on Takens' delay embedding showed the best average accuracy from 69.23 to 76.40%.