Support vector machine（ SVM） is preferably applied to the classification of EEG

and the accuracy rate and time of the classification is determined directly by the parameters of SVM. Applied five kinds of methods of parameters optimization

which are fixed parameters

direct optimization

grid optimization

genetic algorithm optimization

particle swarm optimization. This experiment is tested by BCI competition IV data 2b data sets

the first job is to extract the instantaneous energy features of the data which is processed by Band-Pass filter

secondly which are sent to five kinds of classifiers to optimize

parameter

accuracy rate of nine subjects and time of classification during 4

7 s are obtained. Both of accuracy rates of the classification are highly achieved to96. 875 % through the grid and particle swarm parameters optimization

and the shortest spending time is 3. 059 s.Although the accuracy rate of the classification processed by the direct optimization and fixed parameters is lower

the classification time only needed 0. 002 and 1. 305 s

in terms of real-time

this method is more suitably applied to online system.