Estimation of alertness level by using wavelet transform method and entropy

In this study, developing of a different model estimating of alertness level has been studied by using electroencephalogram (EEG) signals recorded during transition from wakefulness to sleep. Developed model is composed of discrete wavelet transform-entropy pair (feature extractor) and multilayer perceptron neural network (classifier). This study, basically, comprises of two stages. In the first stage, EEG signals taken from 10 healty subjects were separated as alert, drowsy, and sleep signals in the form of 5 s epochs with the aid of expert doctor. In the second stage, feature vectors Delta, Theta, Alpha, and Beta sub-bands of EEG signals separated into epochs were obtained by using discrete wavelet transform. After then, entropy was used to reduce dimensions of feature vectors. Obtained vectors were chosen as input feature vectors of multilayer neural network which used as classifier. Total classification accuracy obtained in the test results of proposed model showed that model can be used in the estimating of vigilance level.