A hybrid BCI study: Temporal optimization for EEG single-trial classification by exploring hemodynamics from the simultaneously measured NIRS data

In this paper we introduced a new method to optimally select the time window for a single-trial classification problem in BCI system. As a hybrid-BCI, we combine EEG and NIRS signals to improve the performance of BCI system. Since there's a coupled relationship between EEG and NIRS, we try to define the activation state of subject's brain according to the changes of hemoglobin. We therefore defined the maximum point of HbO changes to be the time when the brain was fully activated. Then we chose the EEG data according to this critical time point with a 3 s window, which is almost within 6-9s according to the NIRS signal. With this selected time window, there is a significantly improvement of decoding accuracy from 69% to 79% compared to the original time window (1-12 s).