Evaluation of the Hidden Markov Model for Detection of P300 in EEG Signals

Introduction: Evoked potentials arisen by stimulating the brain can be utilized as a communication tool  between humans and machines. Most brain-computer interface (BCI) systems use the P300 component,  which is an evoked potential. In this paper, we evaluate the use of the hidden Markov model (HMM) for  detection of P300.  Materials and Methods: The wavelet transforms, wavelet-enhanced independent component analysis (W- ICA),  and  HMM  combined  with  a  multi-layer  perceptron  (MLP)  neural  network  were  used  for  P300  detection  in  electroencephalogram  (EEG)  signals.  The  BCI2005  competition  dataset  was  used  for  their  evaluation. First, electrooculogram (EOG) artifacts in the EEG signals were removed using W-ICA. Then,  background  EEG  noise  was  suppressed  using  a  B-Spline  wavelet  transform.  Finally,  these  signals  were  classified using the HMM.  Results: We used accuracy, sensitivity, specificity, positive predictive value, and negative predictive value to  evaluate the performance of the proposed algorithm. The primary results in this research show that the HMM  can perform much better using an auxiliary classifier. To this end, an MLP neural network was used to select  the classes based on the outputs of the HMM models. The classification rates obtained for 15 and 5 times  averaged test signals were 81.6% and 50.7% respectively.  Discussion and Conclusion: Based on the obtained results, we may conclude that the HMM can be used for  online P300 detection.