Abstract A single channel EEG signal based sleep stage classification using Discrete Wavelet Transform (DWT) is aimed in this study. DWT is applied to 30-second epochs of the EEG recordings. Recordings from Dreams Project database are used in this study. The EEG signal is filtered by Butterworth low-pass and high- pass filters first. Then, it is decomposed into five sub-bands using DWT according to the American Academy of Sleep Medicine (AASM) standards. Epochs are selected randomly and classified using the presented algorithm. The obtained results are compared with the results scored by an expert of dreams Project internet site. Keywords Sleep stage classification, discrete wavelet transform, EEG signal I. INTRODUCTION LEEP physiological recovery and covering almost one third period of a daytime. A quality and deep sleep is required for efficient regeneration of the body. Sleep stages arise with the evaluation of the quality and deep sleep. EEG signal is commonly used for sleep stage analysis and classification. In literature, the methods for the analysis and classification of EEG based sleep stages are composed of three main steps; (i) Preprocessing of EEG signal (ii) Feature extraction from the EEG signal (iii) Applying extracted features to a classifier Fig. 1 EEG sleep stages classification The block diagram of the three steps is illustrated in Fig. 1. In Pre-processing stage, processes such as filtering the signal from the distortions and normalizing are realized. Important distinctive features of the signal are obtained in TABLE I Erdem Tuncer/ Bahcecik Vocational and Technical Anatolian High School Kocaeli University, Turkey. Emailid: erdemtuncerr@gmail.com Emine Dogru Bolat/ Kocaeli University, Technical Education Faculty, Kocaeli University, Turkey. Email id: ebolat@gmail.com. TECHNIQUES USED FOR SLEEP CLASSIFICATION [1] Author Year Feature Extraction Classification Schmitt,R.B., et al. 1998 Fourier Transform HMM 1 Heiss, J.E., et al. 2002 - Neuro-Fuzzy Subasi,A., et al. 2005 Discrete WT Neural Network Kerkeni. N. 2005 Fourier Transform Neural Network Doroshenkov, L.G., et al. 2007 Fourier Transform HMM Tang, W.C., et al. 2007 HTT+WT SVM Ebrahimi,F., et al. 2008 Wavelet Packet Neural Network Liu,H.J.,et al 2010 Fourier Transform SVM 2 Vatankhah,M., et al. 2010 Discrete WT SVM+NF 4 Ouyang T.,Lu,H.T. 2010 Continuous WT SWM Liu,Y., et al. 2010 HHT 3 Neural Network Le Quoe Khai,Truong Quang Dang Khoa.et[2] 2011 FFT Hierarchical Manner Ms.Vijaylaxmi.P.Jain, Dr.V.D.Mytri. et.al.[3] 2012 Discrete Wavelet Transform Neural Network Guohun Zhu, Yan Li[4] 2013 Mapped into a VG 5 and a HVG 6 SVM Khald A.l.Aboalayon,Helen T.et.al.[5] 2014 Statistical Features Extraction SVM Marwa Obayya F.E.Z.Abou-Chadil[6] 2014 Spectral and Wavelet Analyses Fuzzy C-Means Algorithm 1 Hidden Markov Model 2 Support Vector Machine 3 Hilbert Huang Transform 4 NeuroFuzzy 5 Visibility Graph 6 Horizontal Visibility Graph Feature Extraction Stage. Studies in literature show three main groups of extracted features as given below. 1- Features obtained in time domain 2- Features obtained in the frequency domain 3- Features obtained both in time and frequency domain In the last stage, Classification, the results are obtained using the algorithm based on the extracted features. Some studies about the classification of sleep stages from 1998 up to now are given in TABLE I. Feature Extraction and Classification methods are also stated in this TABLE. In this study, EEG signal is decomposed into sub-bands using discrete wavelet transform. The features of these sub- EEG Signal Based Sleep Stage Classification Using Discrete Wavelet Transform Erdem Tuncer, and Emine Dogru Bolat S Input EEG signal Pre-processing Feature Extraction Classification International Conference on Chemistry, Biomedical and Environment Engineering (ICCBEE'14) Oct 7-8, 2014 Antalya (Turkey) http://dx.doi.org/10.17758/IAAST.A1014055 57