Probabilistic Methods in Multi-Class Brain-Computer Interface

Two probabilistic methods are extended to research multi-class motor imagery of brain-computer interface(BCI):support vector machine(SVM) with posteriori probability(PSVM) and Bayesian linear dis-criminant analysis with probabilistic output(PBLDA).A comparative evaluation of these two methods is conducted.The results shows that:1) probabilistic information can improve the performance of BCI for subjects with high kappa coefficient,and 2) PSVM usually results in a stable kappa coefficient whereas PBLDA is more efficient in estimating the model parameters.