JOURNAL OF TELECOMMUNICATIONS, VOLUME 6, ISSUE 1, DECEMBER 2010 48
ARTEFACTS REMOVAL IN EEG SIGNAL USING WAVELET TRANSFORM AND ARTIFICIAL NEURAL NETWORK Prof Dr. R. Kawitkar and Ms. Rohini More. Abstract— Electroencephalography is a medical imaging technique that reads scalp electrical activity generated by brain structures. The electroencephalogram (EEG) is defined as electrical activity of an alternating type recorded from the scalp surface after being picked up by metal electrodes and conductive media. The EEG measured directly from the cortical surface is called electrocortiogram while when using depth probes it is called electrogram. We will refer only to EEG measured from the head surface. The recognition of epileptic waveform from EEG signal is important physiolgical task as epilepsy is still one of the most frequently occurring disorder. The main goal of this paper is to provide new method to diagnose the epileptic waveform directly from the EEG, by performing quick signal processing which makes it possible to apply in on-line monitoring system. Use of ANN makes the rate of recognition very high and also makes the on-line monitoring and ‘paperless’ task of EEG analysis. Index Terms: Electroencephalography, EEG, Wavelet transform, Artificial Neural Network(ANN).
—————————— ——————————
1 INTRODUCTION EEG signals are the electrical activities in the cortex or on the surface of scalp causing by the physiological activities of the brain. In general, EEG signals are the typical multi-dimensional nonstationary random processes. Classification of the changes of these special waves is critical for understanding of brain functions. EEG was discovered by Hans Berger in 1929. Brain patterns form wave shapes that are commonly sinusoidal. These EEG signals are measured from peak to peak and normally range from 0.5 to 100 V in amplitude, which is about 100 times lower than ECG signals. Brain waves have four main spectral components as (>13 Hz), (8-13 Hz), (4-8 Hz) and (0.5-4 Hz). Epileptic seizures are occational expressions of fundamental and continuous brain disorder. In most cases they appear as high magnitude waves of frequency around 3Hz. So, when these seizures will occur , reflected by waves.
complishes the recognition of epileptic waveforms. Figure-1 shows the detection method.
2 METHODOLOGIES The epileptic waveform detection is done in two steps In the first step by using multi-resolution wavelet decomposition, we obtain different spectral components(α, β, δ, θ) of the measured signal. These components serve as input signals for the artificial neural network(ANN), which ac• Prof. Dr. R.S. Kawitkar : working as a Professor in E&TC Dept. in Sinhgad college of Engineering in Pune, Maharashtra(India)
Fig1: The detection method
2.3 DISCRETE WAVELET TRANSFORM Calculating wavelet coefficients using contnuous wavlet transform is fair amount of work and it will generate an awful lot of data, hence discrete wavelet transform is suggeted.
• Rohini More: Post-graduate student of. Sinhgad college of Engineering in Pune, Maharashtra(India) © 2010 JOT http://sites.google.com/site/journaloftelecommunications/
JOURNAL OF TELECOMMUNICATIONS, VOLUME 6, ISSUE 1, DECEMBER 2010 49
2.2 ANN BASED DETECTION ANN uses standard back propagation feed forward three-layer network having one hidden layer. Input layer has 650 nodes, hidden layer ha 11 nodes and 1 output node, since ANN has to distinguish between epileptical wave and normal wave.
3 GUI FOR EEG ANALYSIS
Fig2: The DWT algorithm Figure 2 illustrates this procedure, where x[n] is the original signal to be decomposed, and h[n] and g[n] are lowpass and highpass filters, respectively. The bandwidth of the signal at every level is marked on the figure as "f". The signal is band limited to 60 Hz using a band stop filter. Then we take 4 level wavelet decomposition. We use ‘sym6’ wavelet because it is smooth wavelet than other wavelet such as sym2,sym3 etc. so reconstruction of EEG signal using various wavelet coefficients is possible with less noise. ‘sym’ wavelet family is modification to ‘db’ wavelet family so we use ‘sym6’ wavelet. ‘symN’ is symmentrical wavelet family where N is order of filter, 2N is length of filter. We get the alpha, beta, delta and theta waves using filters at various decomposition levels. Delta waves are used for detection of Epilepsy so we have taken four level decomposition which gives us signal up to 3.5 Hz only and then delta wave can be easily separated using Butterworth band pass filter having cutoff frequency 1, 3 Hz , order of filter is 10.
Fig4: The GUI ued for analysis
4. RESULT AND CONCLUSION This new method of detection will provide a good accuracy and GUI will help neurologist for on-line analysis of EEG signal. We have tested our tool for 24 data samples out of which 23 are correctly detected. And we have verified these results from neurologist, so our accuracy of detection of disorder is 95.83%
ACKNOWLEDGMENT The authors wish to thank Principal , Head Of The department for their kind support
.
REFERENCES
Fig3: EEG signal and its spectral components after WT
[1] M. Teplan “Fundamentals of EEG Measurement” volume 2, section 2, 2002. [2] James N. Knight “Signal Fraction Analysis and Artifact Removal in EEG” 2003.
50
[3] Minfen Shen, Lisha sun, F.H.Y. Chan “The Classification of Tran sient Time-Varying EEG
Signal via Wavelet Packets Decom
position”. [4] F. S. Tyner, J. R.Knott. 1989. Fundamental of EEG technology, Volume 1: Basic concepts and methods, Raven press, New York. [5] James N. Knight. 2005. Signal Processing And Artifact Removal in EEG Signal, IEEE Trans. Biom. Eng., v. 26, n 6, p. 313-25. [6] The university of Sydney, Fundamentals of Biomedical Engg. Electroencephalogram, notes at http://www.eelab.usyd.edu.au [7] K. J. Blinowska and P. J. Durka, Application of Wavelet Trans form and Matching Pursuit to the Time-Varying EEG Signals, Proc. of Conf. Artif. Neural Networks in Eng., pp.535-540. [8] Engeneering Handbook, pp. 201-212, CRC Press, Florida.J. Pes quet, H. Krim and H. Carfantan, Timeinvariant Orthonomal Wavelet representations.IEEE Trans. on Signal Processing, 44(8): 1964-1970. 1996.
Author: Dr. R. S. Kawitkar: This author has 20 year of experience of teaching. Currently working with Dept. of Electronics and Telecommunication Engineering, Sinhgad College of Engineering, Pune as Professor. He has done Ph.D. (Electronics Engineering) from Amravati University, 2005. MBA (HRM) from YCMOU, (Nashik in 1998).M.S. (Electronics & Control) from BITS-Pilani in 1994 (1st class), B.E. (Electronics) from Amravati University in 1990 (1st class). Ms. Rohini More: This author is a post graduate student of E&Tc engineering, alo has a four year of experience in teaching. Compteted her B.E. from Pune university in 2003 with First Classs & Ditinction. He is a life time member of ISTE Society.
.