- Open Access
- Total Downloads : 17
- Authors : M.Thenmozhi, Ms.K.Subhashini
- Paper ID : IJERTCONV2IS12018
- Volume & Issue : NCACCT – 2014 (Volume 2 – Issue 12)
- Published (First Online): 30-07-2018
- ISSN (Online) : 2278-0181
- Publisher Name : IJERT
- License: This work is licensed under a Creative Commons Attribution 4.0 International License
Elimination of ocular artifacts in EEG signal using ANC and DWT
henmozhi,M.E Communication system/ECE,Sri Sai Ram Engineering College,Chennai, India firstname.lastname@example.org
Ms.K.Subhashini,Assistant Professor/ECE,Sri Sai Ram Engineering College,Chennai, India
Abstract Introduced a new model using DWT and ANC techniques to remove the OAs in contaminated EEG signals. The model is based on discrete wavelet transformation (DWT) and adaptive noise cancellation (ANC). A particularly novel feature of the new model is the use of DWTs to construct an OA reference signal, using the three lowest frequency wavelet coefficients of the EEGs. The results show that the new model demonstrates an improved performance with respect to the recovery of true EEG signals and also has a better tracking performance. The model is also applied and evaluated against data recorded within the EUFP 7 Project Online Predictive Tools for Intervention in Mental Illness (OPTIMI). The results show that the proposed model is effective in removing OAs and meets the requirements of portable systems used for patient monitoring as typified by the OPTIMI project.
Keywords – adaptive noise cancellation; discrete wavelet transformation; electroencephalogram; ocular artifacts;
Nervous disorder is exponentially growing challenge in Tele-care and Tele-health projects. Main example is that EEG, the EEG based home care system in Online Predictive Tools for Intervention in Mental Illness (OPTIMI) which is used to monitor the level of each mental disorders and gather the patients information during treatment using EEG signal.EEG is a recording of the electrical activity of the brain from the scalp. The recorded waveforms reflect the cortical electrical activity. EEG signals are measured from conductors positioned on the scalp. The amplitude scalp ranges from 20 to 200ÂµV.
EEG frequency ranges from 0.1 to 100Hz and its amplitude is 2 to 200ÂµV. EEG consists of five bands; they are delta, theta, alpha and beta. During EEG acquisition, contaminated from eye movements and blinks also produce serious distortion in the recorded data. These environmental factors produce large electric potential around the eyes called ocular artifacts (OAs). Most often used method to remove the ocular artifacts is either time domain  or frequency domain  techniques. Later, principal component analysis ,  is used to remove the artifacts but it cannot remove the artifacts completely from EEG because the waveforms of the ocular artifacts are smaller in amplitude when compared to the ongoing EEG signal.
In the previous research, independent component analysis ,  is used to remove the ocular artifacts but in this analysis the ICA needs reference signal that requires unexciting classification of components , . This techniques is mainly based on wavelet threshold and independent component analysis (ICA) are developed for use in high dimensional neural data. The wavelet techniques uses a discrete wavelet transform with a haar basis function to localize artifacts in time and frequency before removing them with thresholding. In this the wavelet decomposition level is automatically selected based on the smoothness of artifactual wavelet approximation coefficients . The independent component analysis method separates the signal into independent components, detects the artifactual components by measuring the offset between the mead and median of each component, and then removing the correct number of components based on the offset and the power of the reconstructed signal.
A quantitative method for evaluating these techniques is also presented. Through this evaluation, the adaptation of wavelet thresholding to produce admirable reduction of ocular artifacts when compared to the other techniques. In this paper a new model is introduced to remove the artifacts using adaptive noise cancellation and discrete wavelet transform and the most important thing is, we have to construct the reference signal. In this the discrete wavelet transforms to the contaminated signal to derive the reference signal and then it is used in the next stage of ANC. The main advantage of ANC is that it can follows the changes and automatically adjust its parameter ANC is used to remove the noise from the reference signal and often used to remove the power line interference in the EEG signal.
The recorded EEG signals are contaminated by OAs, this contamination is considered to be an additive noise within the EEG signal.
EEGreg (t) = EEGact (t) + k . OAs (t) (1)
EEGreg (t) registered EEG signal EEGact (t) actual EEG signal
k. OAs (t) OAs due to eye movement
Discrete Wavalet Transform
In this model, the wavelet decomposition is used to construct a reference signal and then produce the de-noised signal by the use of adaptive noise cancellation. The de- noised signal is helpful for producing more accurate measurements of latency and time.
Adaptive Noise Cancellation
The noise cancellation in which a process d(n) is to be estimated from a noise corrupted observation
x(n) = d(n) + v1(n) (2)
without any information about d(n) or v1(n) it is impossible to separate the signal from the noise. Fig 1 However, given a reference signal v2(n),that is correlated with v1(n) then this reference signal may be used to estimate the noise v1(n) and this estimate may then be subtracted from x(n) to form an estimate of d(n)
If the reference signal v2(n) is uncorrelated with d(n), then minimizing the mean square error is
equivalent to minimizing .In this
model, the adaptive noise cancellation is mainly used to remove the noise from the reference signal. An alternative method of estimating signals corrupted by additive noise interference is to use an ANC adaptive filter , . In an ANC filter, the interference source is used as a reference when adjusting coefficients automatically to achieve optimal filtering.
LMS and RLS
The core of adaptive filter is the adaptive algorithm From that the most commonly used algorithm is least mean square and recursive least square. In the adaptive system, it is based on LMS algorithm. It has good performance when the environmental noise is a stationary random signal. If the noise is unstable then the LMS has difficult in tracking the noise adaptively as the statistical properties changes.
So, in this model it use RLS algorithm in the adaptive filter because when compared to LMS, the RLS has good performance in a non-stationary environment and is able to track slowly varying parameters. ANC is mainly based on RLS algorithm and it is used to remove the OAs from the EEG signal. The filter adjusts its coefficients in accordance with the weighted square error and minimum standard deviation to get the optimal filter coefficients.
Wn = [ wn(0),wn(1)..wn(p)]T (4)
Fig 1. Adaptive noise cancellation with a reference signal
In that equation it minimize at time n, and the weighted least square error,
Where 0 1 is an exponential weighting factor and e (i) = d (i) y (i) (6)
The stages of our new model are:
Wavelet decomposition is used to decompose signals into multi-scale representations. It is a commonly used tool for analyzing non-stationary signals. The wavelets used in DWT are effective in constructing both time- and fequency-domain information from time-varying and non-stabile EEG signals. Unlike the Fourier transform, the wavelet transform
can use a variety of different basis functions with differing properties. There has been much research on the use of DWT to remove the artifacts in EEG signals.
Fig 2. Block Diagram
According to the minimum risk value, we have to select the soft threshold and apply them to the three lowest level coefficients to get the new coefficient for those levels. The soft threshold is mainly used to make the wavelet coefficients contract to zero according to a fixed vector. The threshold value is given by
Fig 3. EEG signal
decomposition level is too high, the signal to-noise ratio is poor. We, therefore, selected seven layers of decomposition as a compromise. Fig 4. Shows the seven levels of wavelet decomposition for constructing the reference signal.
k = level of decomposition Tk = threshold value
S = number of samples
k= noise signal standard deviation
Here, we apply wavelet reconstruction to the new wavelet coefficients for constructing the reference signal.
ANC based on the RLS algorithm is used to remove the OAs from the EEG signal
RESULT AND DISCUSSION
Thus, after specifying these models the simulated noisy EEG signal is obtained. Each simulated EEG waveform
contains 5120 samples and its sampling frequency is 256 Hz. Fig 3. Shows the actual EEG signal.
The level of the wavelet decomposition is too small, it is difficult to remove the noise effectively; however, if the
Fig 4. Seven levels of wavelet decomposition
Fig 5. Wavelet Decomposition
The accuracy of each model was measured by comparing e(n) with the true EEG x(n) in the time and frequency domains. Here ,we calculate PSNR value by the use of MSE.MSE is given by,
Fig 6. Denoised signal
Where S = 5120 is the total number of samples in each simulated waveform. In frequency-domain analysis, the mean absolute error (MAE) is defined for each of the four
bands: band (f < 4 Hz), band (4 f < 8 Hz), band (8
f < 13 Hz), and band (13 f < 30 Hz)
Where Pe(m) and Px(m) are the power spectral density functions of e(n) and x(n) and then the PSNR value is given by
In the previous model, the PSNR vale is 51.19 db but in this model the PSNR value is 70.4618 db
Thus, the new model using DWT and ANC techniques to remove ocular artifacts in contaminated EEG signals. The above work demonstrates the effectiveness of the new model by using the model to process simulated and standard EEG data. The proposed method eliminates Ocular artifacts in the low frequency band even when their frequency is overlapping with that of the EEG signal. Using simulated data, it has been established that the superior performance of the proposed model does
adversely affect the performance of the model with respect to the application of ICA. Therefore conclude that with respect to other concepts in this area the proposed model is able to provide better attenuation levels for common types of OAs present in EEG signals
In future enhancement, here using Adaptive Neuro Fuzzy Interference System (ANFIS) to find whether the signal is normal or abnormal. ANFIS is a kind of neural network that is based on TakagiSugeno fuzzy inference system. Since it integrates both neural networks and logic principles, it has potential to capture the benefits of both in a single framework. Its inference system corresponds to a set of fuzzy IFTHEN rules that have learning capability to approximate nonlinear functions. Hence, ANFIS is considered to be a universal estimator .Here ANFIS is used for better performance in the peak signal to noise ratio value.
G Gratton, M. G. Coles, and E. Donchin, A new method for off-line removal of ocular artifact, Electroencephalogr. Clin.Neurophysiol, vol. 55, no. 4, pp. 468484, Apr. 1983.
J. C. Woestengurg, M. N. Verbaten, and J. L. Slangen, The removal of the eye movement artifact from the EEG by regression analysis in the frequency domain, Biological Psychol., vol. 16, pp. 127147, Feb./Mar. 1983.
T. D. Lagerlund, F. W. Sharbrough, and N. E. Busacker, Spatial filtering of multichannel electroencephalographic recordings through principal component analysis by singular value decomposition, Clin. Neurophysiol., vol. 14, no. 1, pp. 7382, 1997.
I. T. Joliffe, Principal Component Analysis. New York, NY, USA: Springer-Verlag, 1986.
R. N. Vigario, Extraction of ocular artifacts from EEG using independent component analysis, Electroencephalogr. Clin. Neurophysiol., vol. 103, pp. 395 404, 1997
R. Vigario, J. Sarela, V. Jousmaki, M. Hamalainen, and
E. Oja, Independent component approach to the analysis of EEG and MEG recordings, IEEE Trans. Biomed. Eng., vol. 47, no. 5, pp. 589593, May 2000.
W. Lu and J. C. Rajapakse, ICA with reference, in Proc. 3rd Int. Conf. Independent Component Analysis Blind Signal Separation, 2001, pp. 120125.
A. HyvÂ¨arinen and E. Oja, A fast fixed-point algorithm for independent component analysis, Neural Comput., vol. 9, no. 7, pp. 14831492, 1997
P. L. Feintuch, An adaptive recursive LMS filter,
Proc. IEEE, vol. 64, no. 11, pp. 16221624, Nov. 1976.
M. Rupp, A family of adaptive filter algorithms with decorrelating properties, IEEE Trans. Signal Process., vol. 46, no. 3, pp. 771775, Mar.1998.