Automatic Sleep Stages Detection Based on EEG Signals Using Combination of Classifiers

Sleep stages classification is one of the most important methods for diagnosis in psychiatry and neurology. In this paper, a combination ofthree kinds of classifiers are proposed which classify the EEG signal into five sleep stages including Awake, N-REM (non-rapid eye movement)stage 1, N-REM stage 2, N-REM stage 3 and 4 (also called Slow Wave Sleep), and REM. Twenty-five all night recordings from Physionet database are used in this study. EEG signals were decomposed into the frequency sub-bands using wavelet packet tree (WPT) and a set of statistical features was extracted from the sub-bands to represent the distribution of wavelet coefficients. Then, these statistical features are used as the input to three different classi iers: (1) Logistic Linear classi ier, (2) Gaussian classi ier and (3) Radial Basis Function classi ier. As the results show, each classifier has its own characteristics. It detects particular stages with high accuracy but, on the other hand, it has not enough success to detect the others. To overcome this problem, we tried the majority vote combination method to combine the outputs of these base classifiers to have a rather good success in detecting all sleep stages. The highest classification accuracy is obtained for Slow Wave Sleep as 81.68% in addition to the lowest classi ication accuracy of 43.68% for NREM stage 1. The overall accuracy is 70%.