USING GROUP METHOD OF DATA HANDLING FOR BRAIN COMPUTER INTERFACE

摘要

Electroencephalograms (EEG) contain the information important for brain-computer interface (BCI), despite that their earlier use is limited by variations caused by background brain activity of operators as well as by artifacts. Thus achieving high accuracy of EEG classification is an important task for BCI systems. In this paper, we explore possible ways of improving the classification performance by using Group Method of Data Handling (GMDH). Specifically, we aim to find the most suitable selection criteria to be used in GMDH. The experimental results conducted on EEG representing the motor imagery indicates that the proposed criteria are capable of improving the classification performance as compared with Radial Basis networks.