Subject independent BCI based on LTCCSP method and GA wrapper optimization

Recent BCIs mainly need calibration sessions for a new user in order to system training before the usage. Such systems are known as subject-dependent BCIs which are suitable for just one particular subject. In this research, we proposed an efficient subject-independent BCI that can be applicable for any new subject without the need to calibration session in order to train the BCI system. For this aim, a new approach based on the Local Temporal Correlation Common Spatial Pattern (LTCCSP) method for feature extraction, and GA wrapper algorithm for time interval and frequency band optimization is proposed for designing motor imagery based subject-independent BCIs. According to the experimental results, the suggested Subjectindependent algorithm is able to classify different motor imagery tasks of the new users, efficiently.