A hybrid SVM/HMM classification method for motor imagery based BCI

It is important for Motor Imagery (MI) based Brain-Computer Interface (BCI) system to detect motor states efficiently and accurately. In this study, we presented a hybrid classification method that combined Support Vector Machine (SVM) with Hidden Markov Model (HMM) to improve BCI classification accuracy. The output of SVM was converted into posterior probability acting as the internal hidden state observation probability of HMM. The proposed method was compared with SVM using data sets 2a of the BCI Competition IV. The experiment results show that the proposed hybrid SVM/HMM method outperforms the SVM method and achieves significant accuracy improvement. Moreover, the obtained results also indicate that the hybrid method has a better performance not only in two-class classification tasks but also in multi-class classification tasks. 1553-9105/Copyright © 2015 Binary Information Press