ssMRP state detection for brain computer interfacing using hidden Markov models

This paper reports preliminary results of steady-state movement related potential (ssMRP) classification using hidden Markov models (HMM). Published works on electroencephalogram (EEG) signal classification mainly need experimenter interventions to accurately define temporal boundaries between the resting and motor execution states for the classifier where for asynchronous brain computer interfacing (BCI), the classifier itself should autonomously differentiate between these states. We develop a HMM-based classifier for a three-class BCI problem, i.e. rest, left/right finger tapping. Note that in contrast to [Nazarpour, et.al; 2009], we here experimentally select the best pair of channels which attain the highest classification score instead of the 45 electrodes all over the sensorimotor cortex. The averaged correct classification rates (CCR) for different sliding time windows are reported. Reliable single trial classification rates of approximately 60%-80% accuracy are achievable.