MEV


 

 

Mechatronics, Electrical Power, and Vehicular Technology 05 (2014) 129-138 

 

Mechatronics, Electrical Power, and 

Vehicular Technology 
 

e-ISSN: 2088-6985 

p-ISSN: 2087-3379 

Accreditation Number: 432/Akred-LIPI/P2MI-LIPI/04/2012 

 

 

www.mevjournal.com 
 

 

© 2014 RCEPM - LIPI All rights reserved 

doi: 10.14203/j.mev.2014.v5.129-138 

AN EXPERIMENT OF OCULAR ARTIFACTS ELIMINATION FROM 

EEG SIGNALS USING ICA AND PCA METHODS 
 

Arjon Turnip 
a,
*, Iwan R. Setiawan 

a
, Edy Junaidi

 b
, Le Hoa Nguyen 

c
 

a
Technical Implementation Unit for Instrumentation Development, 

Indonesian Institute of Sciences, Bandung, Indonesia 
b
Indonesian University of Education, Department of Physics, Bandung, Indonesia  

c
Dept. of Electrical Engineering, University of Science and Technology-The University of Danang, Vietnam. 

 
Received 24 April 2014; received in revised form 10 November 2014; accepted 10 November 2014 

Published online 24 December 2014 

 

Abstract 
In the modern world of automation, biological signals, especially Electroencephalogram (EEG) is gaining wide 

attention as a source of biometric information. Eye-blinks and movement of the eyeballs produce electrical signals 

(contaminate the EEG signals) that are collectively known as ocular artifacts. These noise signals are required to be 

separated from the EEG signals to obtain the accurate results. This paper reports an experiment of ocular artifacts 

elimination from EEG signal using blind source separation algorithm based on independent component analysis and 

principal component analysis. EEG signals are recorded on three conditions, which are normal conditions, closed 

eyes, and blinked eyes. After processing, the dominant frequency of EEG signals in the range of 12-14 Hz either on 

normal, closed, and blinked eyes conditions is obtained. 

 

Keywords: EEG, EOG, ICA, PCA, artifacts elimination. 

 

I. INTRODUCTION 
The electrical activity produced by the brain is 

recorded by the electroencephalogram (EEG) 

using several electrodes placed on the scalp due 

to the effect of millions of neurons. Signals 

characteristics vary from one state to another, 

such as wakefulness/sleep or normal. Classically, 

five major brain waves can be distinguished by 

their frequency ranges: delta 0.5–4 Hz, theta 4–8 

Hz, alpha 8–13 Hz, beta 13–30 Hz, and gamma 

30–128 Hz [1].  

The EEG has been developed in various fields 

such as the medical field, the development of 

Brain Computer Interface (BCI) [2-6], etc. In the 

medical field EEG is used to diagnose diseases 

such as epilepsy, metabolic encephalopathy, and 

cerebral parenchyma infective. Applications of 

EEG in the medical field are also able to 

determine the treatment to patients who have 

diseases related to nerves. In the development of 

BCI, EEG signals are used to move the cursor in 

two dimensions [7,8], to design a virtual 

keyboard [9,10], to move the three-dimensional 

simulations [11-14], and to design a mobile robot 

for users who have a physical or mental 

deficiency [2-4]. 

The recorded EEG signals are not only the 

original signal according to the brain activity but 

also contaminated by noise signals such as eye 

blink, eye movement, muscular movement, line 

noise, etc. [1,15-17]. Eye-blinks and movement 

of the eyeballs produce electrical signals that are 

collectively known as Ocular Artifacts (OA). The 

OA are pervasive problems in event-related 

potential (ERP) research.  

The electric potentials created during saccades 

and blinks can be orders of magnitude larger than 

the EEG and can propagate across much of the 

scalp, masking and distorting brain signals. 

Therefore, these noise signals are required to be 

separated from the EEG signals to obtain the 

accurate results. 

In a wide variety of research, artifact removal 

has been done by various methods. Signal 

recording method performed with a variety of 

stimulus, such as moving the limbs, moving the 

eyes, etc. Not only signal recording methods is 
* Corresponding Author. Tel: +62-22-2503053 
E-mail: jujhin@gmail.com 

http://dx.doi.org/10.14203/j.mev.2014.v5.129-138


A. Turnip et al. / Mechatronics, Electrical Power, and Vehicular Technology 05 (2014) 129-138 130 

diverse, signal processing methods are also 

diverse. Among of them is independent 

component analysis (ICA) [15], principal 

component analysis (PCA) [18], adaptive filters, 

autoregressive models [19], non-liniear PCA [20], 

neural networks [21-25], wavelet denoising 

[1,15], gyroscope signal [16] etc. All of methods 

will be more powerfull when the artifacts are 

well identified. 

This paper describes an experiment to identify 

the artifacts with completely automated method 

for eliminating electro ocular contamination from 

EEG signals using statistical criteria applied to 

data components obtained using a blind source 

separation (BSS) algorithm based on independent 

component analysis (ICA) and principal 

component analysis (PCA). 

The artifact is composed of horizontal and 

vertical eye movements, and eye blinks as 

indicated in Table 1. In the experiment, the EEG 

signals when normal condition, closed eyes, and 

blinked eyes are recorded. ICA attempts to 

separate the EEG signals recorded into 

statistically independent sources (components), 

and then reject it with regard to artifacts. PCA is 

widely used for feature extraction by removing 

features that have no significant variance. 

 

II. METHODS 
This experiment involved seven male subjects 

with age range from 20 to 25 year old. All 

subjects had never done recording the EEG 

signals before. EEG signals were recorded using 

the Emotiv Epoch system (Figure 1(a)). EEG 

signals were recorded on a six channel is F7, F8, 

T7, T8, O1, and O2 as they relate to visual 

activity. Electrode placement pattern is shown in 

Figure 1(b).  

The first step of the experiment is the 

electrode preparation. Electrodes are firstly 

smeared using an electrolyte liquid to improve 

conductivity of the electrode. This process takes 

approximately 5 minutes, following by the 

pairing the Emotiv Epoch system on the subject 

and adjust the location of the electrodes on the 

scalp to record. Experiments carried out with 

three stimuli, i.e., normal conditions, closed eyes, 

and blinked eyes. Data recording scenarios is 

shown in Table 1. 

Prior to ocular artifacts elimination using BSS 

based on ICA and PCA, a preprocessing 

operations filter FIR chebyshev type II was 

carried out. Flowchart of signals processing is 

shown in Figure 2. 

 

A. Independent Component Analysis 
ICA is one of algorithms group to solve the 

problem of blind source separation. Independent 

component analysis can be used to estimate     

based on independent information and this makes 

it possible to separate the original signals from 

their mixtures.  

Table 1.  

Data recording scenarios 

Time (second) Stimulus Activity 

0-20 + Normal 

20-40 ↓ Closed eyes 

40-41 ↑ Blinked eyes 

41-45 + Normal 

45-46 ↑ Blinked eyes 

46-50 + Normal 

50-51 ↑ Blinked eyes 

51-55 + Normal 

55-56 ↑ Blinked eyes 

56-60 + Normal 

 

 

    

  (a)  (b) 

 

Figure 1. (a) Emotiv Epoch system; (b) Electrode placement pattern 



 A. Turnip et al. / Mechatronics, Electrical Power, and Vehicular Technology 05 (2014) 129-138 

 
131 

The observed signal is denoted by x with 

elements        , d and the original signal is 
denoted by s with elements        , and A is the 
mixing matrix    . With the vector notation, 

mixing models can be written as follows [20, 21]:  

      (1)  

Estimating the independent components can 

be accomplished by finding the right linear 

combinations of the mixture variables, with 

invert the mixing as follow [20, 21]: 

        (2) 

thus, to estimate one of the independent 

components, we can consider a linear 

combination of the   . Assume a new vector as 

       ∑     (3) 

where b is a vector to be determined [20, 21].  

  ( )       ( )      ( )      ( )  

  ( )       ( )      ( )      ( ) (4) 

  ( )       ( )      ( )      ( )  

by substituting equation (1) to equation (3) it can 

be written       . Thus, y is a certain linear 
combination of the   , where  

   is a coefficient 
matrix denoted by q [20, 21]. So the equation (3) 

can be obtained 

          ∑     (5) 

if b was one of the rows of the inverse of A, this 

linear combination     would actually equal one 

of the independent components. In that case, the 

corresponding q would be such that just one of its 

elements is 1 and the others are zero.  

In practice, b and A cannot be determined but 

we can find an estimator that gives a good 

approximation. The fundamental idea here is that 

since a sum of two independent random variables 

is more Gaussian than the original variables, 

      is usually more Gaussian than any of the 
   and becomes the least Gaussian when it in fact 
equals to one of the   . In this case, obviously 
only one of the elements    of q is nonzero. In 
practice the values of q is unknown, but     
    by the definition of q. Such a vector would 
necessarily correspond to a      , which has 
only one nonzero component. This means that 

          equals to one of the independent 
components [20]. 

 

B. Principal Component Analysis 
The starting point for PCA is a random vector 

x (i.e.,  ( )    ( )) with n elements. Typically 
the elements of x are measurements like pixel 

gray levels or values of a signal at different time 

instants. It is essential in PCA that the elements 

are mutually correlated, and there is thus some 

redundancy in xx. In the PCA transform, the 

vector x is first centered by subtracting its mean 

[20]:  

     * + (6) 

The mean in this practice is estimated from 

the available sample  ( )    ( ). Assume in 
the following that the centering has been done 

and thus  * +   . Next, x is linearly 
transformed to another vector y with m elements 

(m < n) so that the redundancy induced by the 

correlations is removed. This is done by finding a 

rotated orthogonal coordinate system such that 

the elements of x in the new coordinates become 

uncorrelated.  

At the same time, the variances of the 

projections of x on the new coordinate axes are 

maximized so that the first axis corresponds to 

the maximal variance, the second axis 

corresponds to the maximal variance in the 

direction orthogonal to the first axis, and so on. 

In mathematical terms, consider a linear 

combination [20, 21]:  

   ∑      
 
      

 
 , (7) 

where the vector x is the elements        . The 
          are scalar coefficients or weights of 

an n- dimensional vector   , and   
 

 denotes 

the transpose of   . The factor    is called the 
first principal component of x when the variance 

of    is maximally large. Because of the variance 

 

 
 

Figure 2. Flowchart of signals processing 

 



A. Turnip et al. / Mechatronics, Electrical Power, and Vehicular Technology 05 (2014) 129-138 132 

is depends on both the norm and orientation of 

the weight vector   , impose the constraint that 
in the norm of    is constant become equal to 1 
in practice. Thus we look for a weight vector    
maximizing the PCA criterion as follow 

  
   (  )   *  

 +   {(  
 
 )
 
}  

  
 
 *   +     

 
     (8) 

so that ‖  ‖   . The  * +  is the expectation 
over the (unknown) density of input vector x. The 

matrix    is the n x n covariance matrix of x 
given for the zero-mean vector x with the 

correlation matrix as 

    *  
 + (9) 

It is well known from basic linear algebra that 

the solution to the PCA problem is given in terms 

of the unit-length eigenvectors         of the 
matrix   . The ordering of the eigenvectors is 
such that the corresponding eigenvalues 

        satisfy           . The 
solution equation (8) is given by      . Thus 
the first principal component of x is      

  . 

The criterion   
   

 in equation (8) can be 

generalized to m principal components, with m 

any number between 1 and n. Denoting the m-th 

(      ) principal component by    

  
 
 , with    the corresponding unit norm 

weight vector, the variance of    is now 
maximized under the constraint that    is 

uncorrelated with all the previously found 

principal components: 

 *    +        (10) 

note that the principal components    have zero 
means because of:  

 *  +   {  
 
 }     (11) 

The condition (10) yields: 

 *    +   {(  
 
 )(  

 
 )}  

   
 
       (12) 

For the second principal component, we have 

the condition as follow 

  
 
        

 
     (13) 

It is known that      , thus looking for 

maximum variance  *  
 +   {(  

 
 ) } in the 

subspace orthogonal to the first eigenvector of 

  . The solution is given by      . Likewise, 
recursively it follows that      . Thus the k-th 
principal component is      

  . 
 

III. RESULT 
Preparatory to an analysis of the ocular 

elimination from EEG signals, actual signals 

were recorded in a six-channel (F7, F8, T7, T8, 

O1, and O2) configuration as they relate to visual 

activity. The raw data (see Figure 3) were first 

pre-processed using a filter FIR chebyshev type 

 

 

Figure 3. Recorded EEG Signal from the 1
st
 subjects 

 



 A. Turnip et al. / Mechatronics, Electrical Power, and Vehicular Technology 05 (2014) 129-138 

 
133 

II with cut-off frequencies of 0.5 (i.e., to remove 

the trend from low frequency bands) to 49 Hz 

(i.e., to remove unimportant information from 

high frequency bands), respectively (Figure 4). 

The low and the high frequency bands are clearly 

remove in Figure 4 compared to Figure 3 and the 

amplitude of EEG signals turn out to be 20-100 

μV. All channels except channel F8 are highly 

contaminated by medium frequency bands. 

It can be seen that the signals were corrupted 

by noises. Since the original EEG signal to the 

EEG power (noise) ratio is small, a method of 

extracting the brain activity component from the 

EEG is desirable. One way of gaining further 

 

 

Figure 4. Filter EEG Signal using BPF 

 

 

Figure 5. Clean EEG Signal using ICA 

 



A. Turnip et al. / Mechatronics, Electrical Power, and Vehicular Technology 05 (2014) 129-138 134 

insights into EEG signals is by applying BSS 

based on ICA and PCA techniques. Figure 5 and 

Figure 6 shows the feature extraction results of 

the BSS base on ICA and PCA, respectively. The 

visually comparing the time domain plots, it is 

clear that the both algorithm reduces the 

amplitude of the ocular artifact while preserving 

the background EEG. 

The extracted signals using both metods show 

the similarities of the signals propagation in each 

channel with the same condition of stimulus. The 

similarities patterns are given in the red color 

marks. In the closed eyes condition, the 

amplitude is a little bit higher compare with 

others condition. This results indicated that other 

conditions are more contaminated by noise. The 

other results about the highly degradation of the 

amplitude (from 100 μV to 10 μV) also indicated 

in the extracted EEG signals. Compared with the 

ICA method, the extracted EEG signals using 

PCA method produces EEG signals with larger 

amplitude as shown in Figure 6. This indicates 

that the signal is not completely separated from 

the mixture. Although the signals were still 

corrupted by noises (manifested as the high 

amplitudes of the artifacts in some sessions), the 

behaviors of the extracted signals clearly 

represented the brain activity components. 

In order to show the visually performance of 

the extracted EEG signals, the brain mapping 

process is applied. The brain maps using both 

methods is given in Figure 7. The magnitude 

around 500 ms after given stimulus of each 

extracted signals are ploted in 2-D maps. The 

active brain regions that have been separated on 

any channels are indicated with a yellow to red. 

The red color indicated the higher brain activity 

and the blue color indicate the lower brain 

activity. The brain maps using PCA method 

indicates that the ocular artifacts are not perfectly 

separated (specially on channels F8 and O1) 

compared with the ICA method. 

 

 

Figure 7. (a) Brain maps using ICA, (b) Brain maps using 

PCA 

 

 

Figure 6. Clean EEG Signal using PCA 

 



 A. Turnip et al. / Mechatronics, Electrical Power, and Vehicular Technology 05 (2014) 129-138 

 
135 

Other results about average EEG spectra 

based on three conditions stimulus (i.e. normal, 

closed, and blinked eyes condition) in the 

experiment using both methods are given in 

Figure 8 and Figure 9, respectively. The higher 

amplitude using both method is achieved in the 

range frequencies of 12-14 Hz. This results 

indicate that both method are success in eliminate 

the ocular artifacts without losing of important 

information on the recorded EEG signals. 

The ICA method presents the amplitude: 

normal and blinked eyes condition about 1-1,5 

μV while closed eyes condition about 5 μV. The 

PCA method present the amplitude: normal and 

blinked eyes condition about 1.3-1.4 μV while 

closed eyes condition about 18 μV. The 

differences in the level of amplitude between 

both methods indicate the ability of the 

separation of the noise from the brain activity in 

the recorded EEG signals. Dominant frequency 

 

 

Figure 8. Average EEG spectra of each channels on normal, closed, and blinked eyes condition extracted using ICA 

 

 

Figure 9. Average EEG spectra of each channels on normal, closed, and blinked eyes condition extracted using PCA 



A. Turnip et al. / Mechatronics, Electrical Power, and Vehicular Technology 05 (2014) 129-138 136 

and amplitude of each condition for all subjects 

are shown in Table 2 and Table 3 using ICA and 

PCA methods, respectively. 

Table 2 and Table 3 show the brain waves 

spectrum of all subjects (S) in the frequency 

range about 11-16 Hz. There are two subjects 

with high frequency range under normal and 

blinked eyes conditions that are the 4
th

 and 6
th
 

subject which is in the frequency range about 21-

28 Hz. There are two possible reasons for these 

results: the subject was not focus on the given 

stimulus and or was not relaxed during the 

experiment. Overall results of the signal 

processing show the average frequency of normal 

condition, closed eyes, and blinked eyes more 

dominant in the frequency ranges of 12-14 Hz 

(alpha-beta). When the brain on alpha waves, it 

shows the subjects in a relaxed state, and beta 

waves showed that subjects in a state of full 

awareness, it is accordance with the current state 

of the recording signal. There are some subjects 

that have a larger frequency (high beta) than the 

other. When subjects are at high beta conditions, 

it indicate that the subjects were thinking on 

other activities or not to focus on the given 

stimulus. 

 

IV. CONCLUSION 
An experiment for the elimination of eye 

blink artifact from EEG signal using blind source 

separation algorithm based on independent 

component analysis and principal component 

analysis is reported. EEG signals are recorded on 

three conditions, which are normal conditions, 

closed eyes, and blinked eyes. After processing, 

the higher amplitude of EEG signals in the range 

of 12-14 Hz either on normal, closed, and blinked 

eyes conditions is obtained. Both methods are 

successfully eliminate the ocular artifacts without 

losing of important information on the recorded 

EEG signals.  

 

ACKNOWLEDGMENT 
This research was supported by the thematic 

program (No. 3425.001.013) through the 

Bandung Technical Management Unit for 

Instrumentation Development (Deputy for 

Scientific Services) and the competitive program 

(No. 079.01.06.044) through the Research Center 

for Metalurgy (Deputy for Earth Sciences) 

funded by Indonesian Institute of Sciences, 

Indonesia. 

 

 

 

Table 2.  

Frequency (Hz) and brainwave amplitude (µV) in normal, closed, and blinked eyes conditions using ICA 

S 
Normal Closed Eyes Blinked Eyes 

F (Hz) A (µV) F (Hz) A (µV) F (Hz) A (µV) 

1 13 1,143 13 5,029 13 1,575 

2 15 0,918 11 2,691 14 1,035 

3 16 1,290 11 4,001 16 1,014 

4 27 0,935 12 1,015 15 0,725 

5 13 0,839 12 4,763 13 0,912 

6 22 1,017 12 4,948 19 0,739 

7 13 0,520 12 5,830 13 0,371 

 

 

Table 3.  

Frequency (Hz) and brainwave amplitude (µV) in normal, closed, and blinked eyes conditions using PCA 

S 
Normal Closed Eyes Blinked eyes 

F (Hz) A (µV) F (Hz) A (µV) F (Hz) A (µV) 

1 13 1,355 13 18,087 14 1,318 

2 14 1,603 11 6,488 13 2,326 

3 16 0,468 12 2,721 12 0,312 

4 13 1,496 12 9,911 13 1,323 

5 12 4,144 12 14,899 12 3,063 

6 11 0,475 12 14,538 9 0,465 

7 12 4,296 12 17,706 12 4,070 

 

 



 A. Turnip et al. / Mechatronics, Electrical Power, and Vehicular Technology 05 (2014) 129-138 

 
137 

REFERENCES 
[1] R. R. Vázquez, H. V. Pérez, R. Ranta, V. D. 

Louis, D. Maquin, and L. Maillard, “Blind 

source separation, wavelet denoising and 

discriminant analysis for EEG artefacts and noise 

cancelling,” Biomedical Signal Processing and 

Control, vol. 7, pp. 389–400, 2012. 

[2] L. Bi, X.-A. Fan, and Y. Liu, “EEG-Based 
Brain-Controlled Mobile Robots: A Survey,” 

IEEE Transactions On Human-Machine Systems, 

vol. 43, no. 2, pp. 161-176, 2013. 

[3] P. F. Dieza, S. M. T. Müller, V. A. Mut, E. 
Laciar, E. Avila, T. F. B. Filho, and M. S. Filho, 

“Commanding a robotic wheelchair with a high-

frequency steady-state visual evoked potential 

based brain–computer interface,” Medical 

Engineering & Physics, vol. 35, pp. 1155–1164, 

2013. 

[4] A. C. Lopes, G. Pires, and U. Nunes, “Assisted 
navigation for a brain-actuated intelligent 

wheelchair,” Robotics and Autonomous Systems, 

vol. 61, pp. 245–258, 2013.  

[5] A. Turnip, D. Soetraprawata, and D. E. 
Kusumandari, "A Comparison of EEG 

Processing Methods to Improve the Performance 

of BCI,” International Journal of Signal 

Processing Systems, vol. 1, No. 1, 2013. 

[6] A. Turnip, and D. E. Kusumandari, 
“Improvement of BCI performance through 

nonlinear independent component analisis 

extraction,” Journal of Computer, vol. 9, no. 3, 

pp. 688-695, March 2014. 

[7] B. Z. Allisona, C. Brunner, C. Altstätter, I. C. 
Wagner, S. Grissmann, and C. Neuper, “A 

hybrid ERD/SSVEP BCI for continuous 

simultaneous two dimensional cursor control,” 

Journal of Neuroscience Methods, vol. 209, pp. 

299– 307, 2012.  

[8] Y. Li, J. Long, T. Yu, Z. Yu, C. Wang, H. 
Zhang, and C. Guan, “An EEG-Based BCI 

System for 2-D Cursor Control by Combining 

Mu/Beta Rhythm and P300 Potential,” IEEE 

Transactions On Biomedical Engineering, vol. 

57, no. 10, pp. 2495-2504, 2010. 

[9] L. Mayaud, S. Filipe, L. Pétégnief, O. 
Rochecouste, and M. Congedo, “Robust Brain-

computer Interface for virtual Keyboard 

(RoBIK): Project results” IRBM, vol. 34, pp. 

131–138, 2013. 

[10] R. B. Ashari, I. A. Al-Bidewi, and M. I. Kamel, 
“Design and simulation of virtual telephone 

keypad control based on brain computer interface 

(BCI) with very high transfer rates,” Alexandria 

Engineering Journal, vol. 50, pp. 49–56, 2011. 

[11] A. S. Royer, A. J. Doud, M. L. Rose, and B. He, 
“EEG Control of a Virtual Helicopter in 3-

Dimensional Space Using Intelligent Control 

Strategies,” IEEE Transactions On Neural 

Systems And Rehabilitation Engineering, vol. 18, 

no. 6, pp. 581-589, 2010. 

[12] A. Akce, M. Johnson, O. Dantsker, and T. Bretl, 
“A Brain–Machine Interface to Navigate a 

Mobile Robot in a Planar Workspace: Enabling 

Humans to Fly Simulated Aircraft With EEG,” 

IEEE Transactions On Neural Systems And 

Rehabilitation Engineering, vol. 21, no. 2, pp. 

306-318, 2013. 

[13] G. G. Gentiletti, J. G. Gebhart, R. C. Acevedo, 
O. Yá˜nez-Suárez, and V. Medina-Ba˜nuelos, 

“Command of a simulated wheelchair on a 

virtual environment using a brain-computer 

interface,” IRBM, vol. 30, pp. 218–225, 2009. 

[14] H.-J. Hwang, K. Kwon, and C.-H. Im, 
“Neurofeedback-based motor imagery training 

for brain–computer interface (BCI),” Journal of 

Neuroscience Methods, vol. 179, pp. 150–156, 

2009. 

[15] G. Geetha, and S. N. Geethalakshmi, “Artifact 
Removal from EEG using Spatially Constrained 

Independent Component Analysis and Wavelet 

Denoising with Otsu’s Thresholding Technique,” 

Procedia Engineering, vol. 30, pp. 1064 –1071, 

2012. 

[16] S. O’Regan, S. Faul, and W. Marnane, 
“Automatic detection of EEG artefacts arising 

from head movements using EEG and gyroscope 

signals,” Medical Engineering & Physics, vol. 35, 

pp. 867– 874, 2013. 

[17] S. O’Regan, and W. Marnane, “Multimodal 
detection of head-movement artefacts in EEG,” 

Journal of Neuroscience Methods, vol. 218, pp. 

110–120, 2013. 

[18] T. Liu, and D. Yao, “Removal of the ocular 
artifacts from EEG data using a cascaded spatio-

temporal processing,” computer methods and 

programs in biomedicine, vol. 83, pp. 95–103, 

2006. 

[19] V. Lawhern, W. D. Hairston, K. McDowell, M. 
Westerfield, and K. Robbins, “Detection and 

classification of subject-generated artifacts in 

EEG signals using autoregressive models,” 

Journal of Neuroscience Methods, vol. 208, pp. 

181–189, 2012. 

[20] A. Cichocki and S.-I. Amari, Blind Decorrelation 
and SOS for Robust Blind Identification. 

Adaptive Blind Signal and Image Processing, 

Chapter 4, (2002). 

[21] A. Turnip, K.-S. Hong, and M.-Y. Jeong, "Real-
time feature extraction of P300 component using 

adaptive nonlinear principal component 

analysis," Journal of BioMedical Engineering 

OnLine, 10:83, 2011. 

[22] D. Soetraprawata, and A. Turnip, 
"Autoregressive Integrated Adaptive Neural 

Networks Classifier for EEG-P300 

Classification," Journal of Mechatronics, 

Electrical Power and Vehicular Technology, vol. 

4, no. 1, pp. 1-8, 2013. 

[23] H.-A. T. Nguyen, J. Musson, F. Li, W. Wang, G. 
Zhang, R. Xu, C. Richey, T. Schnell, F. D. 

McKenzie, and J. Li, “EOG artifact removal 



A. Turnip et al. / Mechatronics, Electrical Power, and Vehicular Technology 05 (2014) 129-138 138 

using a wavelet neural network,” 

Neurocomputing, vol. 97, pp. 374–389, 2012. 

[24] A. Turnip, and D. Soetraprawata, "Performance 
of EEG-P300 Classification Using 

Backpropagation Neural Networks," Journal of 

Mechatronics, Electrical Power and Vehicular 

Technology, vol. 4, no. 2, pp. 81-88, 2013. 

[25] A. Turnip, and K.-S. Hong, "Classifying mental 
activities from EEG-P300 signals using adaptive 

neural network," International Journal of 

Innovative Computing, Information and Control 

(IJICIC), vol. 8, no. 7, 2012.