Hidden Markov Model (HMM) was evaluated for P300 detection in electroencephalogram (EEG) signal. In some applications like the brain-computer interface (BCI), where real time detection is a concern, HMM could be a useful tool. Wavelet enhanced independent component analysis (wICA) was used for electrooculogram (EOG) artifact removal and B-spline wavelet transform for background EEG noise cancellation. HMM results are enhanced by a multilayer perceptron (MLP) neural network. Accuracy of the proposed HMM classifier is 81.6% on the validation dataset.