Abstract— Electrooculographic (EOG) artefact is one of the most common contaminations of Electroencephalographic (EEG) recordings. The corruption of EEG characteristics from Blinking Artefacts (BAs) affects the results of EEG signal processing methods and also impairs the visual analysis of EEGs. In this paper, our scope was a comparative analysis of the performance of three standard denoising methods like continuous Empirical Mode Decomposition (EMD), Discrete Wavelet Transform (DWT) and Kalman Filter (KF). In order to evaluate the performance of EMD, DWT and KF of noise reduction and to express the quality of the denoised EEG, we calculate several indexes such as the Signal-to-Noise Ratio (SNR). All the results obtained from noise simulated EEG data show that WT achieved the greatest SNR difference and also the mode mixing issue of EMD affected this method’s performance. I. INTRODUCTION Electroencephalogram (EEG) is a noninvasive measurement of the brain’s electrical activity obtained using several electrodes placed on the scalp. Over the last decades and under an increasing medical demand, EEG became an important diagnostic tool for monitoring and managing dysfunctions and various neurological disorders of the human brain. One of the most tempting problems in biomedical signal processing is the extraction of high resolution EEG from contaminated recordings. An increasing number of denoising techniques have been proposed for solving this problem [1]. It is still a challenge to get qualitative or quantitative EEG analysis because of various noise sources that make the denoising process extremely difficult [1]. C. I. Salis is with the Department of Informatics and Telecommunications Engineering, University of Western Macedonia, Kozani GR 50 100, Greece (e-mail: st0347@icte.uowm.gr). A. E. Malissovas is with the Department of Informatics and Telecommunications Engineering, University of Western Macedonia, Kozani GR 50 100, Greece (e-mail: tmalissovas@gmail.com). P. A. Bizopoulos is with the Biomedical Engineering Laboratory, School of Electrical and Computer Engineering, National Technical University of Athens (NTUA), GR-157 73 Zografou, Athens, Greece (e-mail: bizopoulos.paschalis@gmail.com). A. T. Tzallas is with the Department of Informatics & Telecommunications Technology Technological Educational Institute of Epirus, Arta, Greece (e-mail: atzallas@cc.uoi.gr). P. A. Angelidis is Assoc. Prof., Department of Informatics and Telecommunications Engineering, University of Western Macedonia, Kozani GR 50 100, Greece (e-mail: paggelidis@uowm.gr). D. G. Tsalikakis is with the Department of Informatics and Telecommunications Engineering, University of Western Macedonia, Kozani GR 50 100, Greece (corresponding author; e-mail: dtsalikakis@uowm.gr). Electromyographic (EMG), electrodermal response, eye blinks, eye movements, and respiratory are the most common biologically noise sources that generate EEG artefacts [2]. EOG artefacts are one of the main problems of EEG analysis, since EEG recordings are usually contaminated from eye movements and Blinking Artefacts (BAs) which are impossible to prevent. It is essential to estimate EOG accurately, in order to subtract it from the contaminated EEG. The difficulty of achieving this is due to the high amplitude and the low frequency components of the EOG that overlap the frequencies of EEG [2]. Kalman Filter (KF) is not a recently implemented method and it has been employed for EOG detection [3], correction [4] and BA removal [5] with promising results. Wavelets were introduced in early 90s with numerous applications in EEG processing. EEG is a non-stationary signal and several studies, using wavelet adaptive thresholding algorithms, have been applied in order to identify and remove EOG [6]. Also, wavelets have been used as a method for detecting epileptic spikes [7] and denoising Electrocardiographic (ECG) [8]. The Empirical Mode Decomposition (EMD) was introduced as a data driven method for decomposing the signal in components called Intrinsic Mode Functions (IMFs). EMD is a modern adaptive method for detecting and separating the EOG artefacts from EEG signals, with several modifications [9] and combinations with other methods [10]. In this work, the performance of these three methods EMD, WT and KF is quantitatively compared in removing EOG artefacts with different amplitudes from simulated EEG. In order to obtain the denoising results we apply the classic EMD [11], the KF with some modifications [5] and the Discrete Wavelet Transform (DWT) [12, 13]. The efficiency of EMD, WT and KF in rejecting the EOG artefacts was evaluated by calculating several metrics. The results were compared between the contaminated and original signals and between EOG clean and original signals. II. METHODS A. Database Construction In this paper, simulated ΕΕG signals were generated by an algorithm described before [9]. Every signal contains 10000 samples with a sampling frequency 2kHz. Denoising Simulated EEG Signals: A Comparative Study of EMD, Wavelet Transform and Kalman Filter Christos I. Salis, Anastasios E. Malissovas, Paschalis A. Bizopoulos, Alexandros T. Tzallas, P. A. Angelidis and Dimitrios G. Tsalikakis 978-1-4799-3163-7/13/$31.00 ©2013 IEEE