Feature extraction and classification of Electroencephalogram signals for vigilance level detection

Brain Computer Interface uses brain power in the form of Electroencephalogram signals to establish an artificial communication pathway between human brain and outside world. These Electroencephalogram signals alter with the different vigilance levels of human brain. Medicines with high alcoholic contents make patient feel drowsy. This can cause change in the pattern of Electroencephalogram signals recovered from patient and wrong interpretation by classifier algorithm, if change in signals is severe. Further a wrong command can be generated by Brain Computer Interface. In present work, a methodology for feature extraction and classification of EEG signals recorded under drowsy and controlled state of mind is proposed for vigilance level detection of human body. Filtered EEG data is transformed to the time frequency domain and further processed to derive initial parameters based on dynamic programming for nonlinear fitting, to prepare feature vector from raw Electroencephalogram signals. Classifier is trained with the input feature vector and tested with the unseen data. For classification of signals Random Forest Tree classifier is employed.