Motor Imaginary Brain Response Classification using Ensemble Classification Method

Bosun Hwang, Cheolsu Park, and Byung-Tak Zhang

Abstract

Motor imagery brain response during motor planning is one of the most popular paradigms to implement brain-computer interface (BCI) system. The brain response during motor imagery for BCI is commonly obtained using electroencephalogram (EEG) owing to its noninvasive and convenient way to record. The information of interest in EEG is located in well-defined frequency bands, and a number of standard algorithms have been used for feature extraction and pattern classification. We show that factor analysis and ensemble classification methods can be applied to enhance the pattern classification rate to separate two different motor imagery tasks instead of conventional single classification methods. Comparative study on both synthetic benchmark examples and well established BCI motor imagery dataset supports the analysis. (BCI competition IV dataset I & hysiobank motor/mental imagery database)