MPA EEG model-based vigilance level estimation by artificial neural network

In this paper, the vigilance levels during day time short nap sleep were estimated on the basis of Markov Process Amplitude (MPA) EEG model. The ultimate purpose was to adopt the MPA model to discriminate three levels of vigilance through a simple neural network. A set of parameters were firstly calculated based on MPA EEG model. Secondly, correlation analysis was adopted to extract effective parameters to ensure a small amount of inputs of the artificial neural network. The outputs of artificial neural network were the classified three levels: wakeful, drowsy and sleep. The obtained estimation result showed that the accuracy of wakeful was about 90.0%, drowsy 80.0%, and sleep 93.3%.