Classification of single trial EEG during motor imagery based on ERD

EEG-based brain computer interface (BCI) provides a completely new communication channel between human brain and computer. Classification of EEG signals is a difficult task, especially when the classification has to be preformed on a single-trial EEG to continuously control a device. Event related desynchronization (ERD) has proven to be induced on the contralateral sensorimotor area during imagination of a left or right hand movement. In this paper, we introduced a quantification of ERD, with which a lower classification error rate and a higher information transfer rate can be obtained. The performance was tested by the Graz dataset for BCI competition 2003.