HMM based predictive model of brain computer interface

This paper, for effective interaction between user's brain and computer, proposes a Hidden Markov Modelbased prediction approach wherein based on its current state of action, the system calculates the possible outcomes that would lead to the next state/action generated from Hidden Markov Models themselves. These Hidden Markov Models are trained through HMM Toolkit using the frequency features extracted from input EEG waves in the training phase. In the data prediction phase, three sets of ten input EEG waves for different tasks obtained from the end user are compared with the actual training wave data and next state of action that user wants to perform is predicted based on the probability distribution over the possible output tokens of Hidden Markov Models from the training phase. For e.g. for task in which user thinks of opening a music player, on basis of this EEG wave, wave corresponding to playing a song is predicted and system performs it on its own.