Mean Threshold and ARNN Algorithms for Identification of Eye Commands in an EEG-Controlled Wheelchair

doi:10.4236/eng.2013.510B059

Paper Menu >>

Journal Menu >>

Engineering, 2013, 5, 284-291

http://dx.doi.org/10.4236/eng.2013.510B059 Published Online October 2013 (http://www.scirp.org/journal/eng)

Mean Threshold and ARNN Algorithms for Identification

of Eye Commands in an EEG-Contr olled Wheelchair

Nguyen Thanh Hai1, Nguyen Van Trung2, Vo Van Toi1

1Biomedical Engineering Department, International University, Vietnam National University, Ho Chi Minh City, Vietnam

2Faculty of Ele ctrical-Electronics Engineering, University of Technical Education, Ho Chi Minh City, Vietnam

Email: [email protected]

Received July 2013

ABSTRACT

This paper represented Autoregressive Neural Network (ARNN) and meant threshold methods for recognizing eye

movements for control of an electrical wheelchair using EEG technology. The eye movements such as eyes open, eyes

blinks, glancing left and glancing right related to a few areas of human brain were investigated. A Hamming low pass

filter was applied to remove noise and artifacts of the eye signals and to extract the frequency range of the measured

signals. An autoregressive model was employed to produce coefficients containing features of the EEG eye signals. The

coefficients obtained were inserted the input layer of a neural network model to classify the eye activities. In addition, a

mean threshold algorithm was employed for classifying eye movements. Two metho ds were compared to find the better

one for applying in the wheelchair control to follow users to reach the desired direction. Experimental results of con-

trolling the wheelchair in the indoor environment illustr ated the effectiveness of the proposed approaches.

Keywords: Autoregre ss ive NN Mo de l ; Threshold algorithm; EEG Technology; Eye Activity and Electrical

Wheelchair

1. Introduction

Human brain plays an important role in controlling all

body activities [1]. Moreover, it is a complex structure,

in which there are about around 100 billion neurons

which communicate from one to another with or without

external excitations to make control decisions (cognition,

motion, pattern recognition, etc.). For these reasons, non-

invasive technologies such as EEG, functional Magnetic

Resonance Imaging (fMRI) and functional Near-infrared

Spectroscopy (fNIRS) have been investigated to quantify

motor proc e s s ing funct i on of huma n bra in [2-4]. Thus the

exploration of these technologies can allow us to perform

rehabilitative problems or brain simulator leading to im-

prove or recover the motor/cognitive functions of tetrap-

legic patients with spinal cord injuries and degenerative

nerve diseases.

In recent years, EEG technology has quickly devel-

oped and also attracted many researchers related to hu-

man brain. Many Brain-computer Interface (BCI) appli-

cations as well as Brain-based diagnoses have been suc-

cessfully represented, in which BCI problems have been

investigated to implement on human in recent years. In

particular, a BCI system can allow people to communi-

cate and to control external devices [5-7]. It means that

one can translate brain activities into messages or com-

mands to control devices [8-10]. Blankertz et al. devel-

oped the non-invasive BCI system, in which the key fea-

tures were considered to predict the laterality of upcom-

ing left vs. right hand movements to produce the result of

very high information transfer rate.

An EEG system has been used to measure delta signal

of human brain corresponding to eye blinks [11]. However,

to determine the problem of eye activities, a threshold

algorithm was employed to control electric wheelchair.

In this paper, we develop a neural network model [12-14],

in which the inputs of the network are AR coefficients

[15] which were determined based on the filtered EEG

signals using a Hamming lowpass filter from the identi-

fied outputs of the network. In addition, a threshold algo-

rithm is employed to find the m ean thresholds for eye move-

ments. In this res earch, two of these methods will be com-

pared to determine the best one to control the wheelchair.

2. Materials and Methods

2.1. Data Acquisition

Data at Fp1, F7, F8 areas of human brain were obtained

from an Active-Two system as shown in Figure 1. Nine

subjects (males and females, average age: 22 ± 5.33)

were invited to participate into this study. The subjects

informed consents agreement after reading and under-

standing of the experiment protocol and the EEG tech-

N. T. HAI ET AL.

285

Figure 1. Subject with electrodes for obtaining data on the

system.

nique. Offline data were obtained at some positions such

as at Fp1, F 7, F8, CMS and DR L (see Figure 2) on h ead

of each subject through the Active-Two system. The

subjects were instructed to perform their eye activity

times (opening eyes, blinking two eyes, glanced left,

glanced right), each subject performed his/her eye ac-

tivity in 5 seconds.

2.2. Signal Pre-Processing

In the EEG signal processing, the original signal is

passed through a band-pass filter with an impulse re-

sponse in ord er to produce the output of the filter ,

][ ng

For the convolution operation between the EEG signal

and the impulse response of the Hamming low pass filter,

it is described as follows:

[][][][] []

gnxnh nxkhnk

∞

=−∞

=×= −

∑

(1)

where

][ nx

is the EEG signal and ][nhH is the im-

pulse response,

kn,

= 1,2,…N.

The impulse response of the actual Hamming filter is

calculated as follows:





≤≤

=otherwise

N-n nwnh

nhH0

10][][

][

(2 )

where

][nw

is the Hamming window and

][nh

de-

notes the ideal impulse response.

To reject influence by voltage drift, the output signal is

calcula te d using the f ollowing formula:

ngny

∑

−=

][

][][

(3)

In this paper, the number of the EEG signal samples is

N = 1024. The original signal and the filtered signal at

position of F8 are processed as shown in Figure 3.

After filtering noise by the filter, th e filtered EEG sig-

nals corresponding to eye events such as opening eyes,

blinking eyes, glancing left and glancing right (see Fig-

ures 4 and 5) are calculated to determine coefficients for

the recognition of eye activities using neural networ ks.

Figure 2. Five electrodes were installed at 5 positions.

Figure 3. Original signal (F8) and the filtered noisy signal.

2.3. AR Model for Feature Extraction

In this paper, an Autoregression (AR) model is used to

extract the features of the EEG signal. In the A R model,

coefficients are determined using the equation:

)2()1()(

−+−= nyanyany

(4)

where

)(ny

is the filtered EEG signal, n = 1, 2, …,

and

are two coefficients of the AR model.

In the EEG signals collected at three channels FP1, F7,

F8, each channel has two AR coefficients. H ence in four

experiments (blinked, opened eyes, glanced left and

glanced right), one will create four vectors and each vec-

tor has 6 AR coefficients as shown in Table 1.

2.4. Neural Network Model

Classification is an important step to determine the acti-

vity of the eye. After being extracted the signal features

of the eye activities using the AR mode, it produces the

coefficients, which will be transmitted directly into back

propagation neural networks with two hidden layers (see

Figure 6) for training [15].

The back-propagation network is to minimize the erro r

function in the weight space by the reduced gradient me-

thod. Because this method of calculating the gradient of

the error function at each iteration requires that the error

function should be continuous and indivisible. One of the

activation function used in this paper is the sigmoid

CMS

DRL

Fp1 electrode

00.5 11.5 22.5 33.5 44.5 5

-500

500 Si gnal channel F 8 (Raw)

Ti me (s)

Amplitude

00.5 11.5 22.5 33.5 44.5 5

-50

100

150 Si gnal channel F 8 (Fi l tered)

Ti me (s)

Amplitude

N. T. HAI ET AL.

286

Table 1. Vector of ar coefficients for four experiments.

Opening eyes (ao) Blinking eyes (ab) Glancing left (al) Glancing right (ar)

Fp1 F7 F8 Fp1 F7 F8 Fp1 F7 F8 Fp1 F7 F8

ao11 ao71 ao81 ab11 ab71 ab81 al11 al71 al81 ap11 ar71 ar81

ao12 ao72 ao82 ab12 ab72 ab82 al12 al72 al82 ap12 ar72 ar82

(a)

(b)

Figure 4. (a) Filtered signal y[n] in case of opening eyes; (b)

Filtered signal y[n] in case of blinking eyes .

function, which is described as follows:

xS −

)(

(5)

Consider a back-propagation neural network with n

input, m output, contains a number of hidden layer neu-

rons to form the training data set (on - off), in which the

desired set, (x1,d1), (x2,d2),..., (xp,dp) contains P m × n pair

of vectors. The weights will be chosen at random. When

the data set xi are trained to create the different outpu t set

(Oi,di), then the error function E is calculated by the fol-

lowing f ormula:

∑

−=

dOE

)(

(6)

where P is the number of samples, O is the network out-

put, d de n otes the de s i r ed output and

is constant.

(a)

(b)

Figure 5. (a) Filtered signal y[n] in case of glancing left; (b)

Filtered signal y[n] in case of glancing right.

Figure 6. The structure of a NN model with two hidden

layers.

The back-propagation algorithm is used to find the lo-

cal minima of the error function. Therefore, the gradient

of the error function is calculated to change the initial

weight values for the network. The weights are the pa-

rameters changed to reduce errors and then each weight

will increase a typical value:

w∂

∂

−=∆

(7)

00.5 11.5 22.5 33.5 44.5 5

-50

50 Si gnal channel F p1 (Fi l tered)

Ti me (s)

Ampl itude

00.5 11.5 22.5 33.5 44.5 5

-100

100 Si gnal channel F 7 (Fi l tered)

Ti me (s)

Ampl itude

00.5 11.5 22.5 33.5 44.5 5

-50

50 Si gnal channel F 8 (Fi l tered)

Ti me (s)

Ampl itude

00.5 11.5 22.5 33.5 44.5 5

-500

500 Si gnal channel F p1 (Fi l tered)

Ti me (s)

Amplitude

00.5 11.5 22.5 33.5 44.5 5

-200

200 Si gnal channel F7 (F i l t ered)

Ti me (s)

Amplitude

00.5 11.5 22.5 33.5 44.5 5

-200

200 Si gnal channel F8 (F i l t ered)

Ti me (s)

Amplitude

00.5 11.5 22.5 33.5 44.5 5

-100

100 Si gnal channel F p1 (F il tered)

Ti me (s)

Amplitude

00.5 11.5 22.5 33.5 44.5 5

-100

100 Si gnal channel F 7 (F il tered)

Ti me (s)

Amplitude

00.5 11.5 22.5 33.5 44.5 5

-200

200 Si gnal channel F 8 (F il tered)

Ti me (s)

Amplitude

00.5 11.5 22.5 33.5 44.5 5

-100

100 Si gnal channel Fp1 (F i l t ered)

Ti me ( s)

Amplitude

00.5 11.5 22.5 33.5 44.5 5

-200

200 Si gnal channel F7 (F i l t ered)

Ti me ( s)

Amplitude

00.5 11.5 22.5 33.5 44.5 5

-200

200 Si gnal channel F8 (F i l t ered)

Ti me ( s)

Amplitude

•

Input

Layer

Hidden-1

Layer

Output

Layer

Hidden-2

Layer

N. T. HAI ET AL.

287

where the weight vector w is in the network,

denotes

the learning rate (constant),

wE ∂∂ /

is the derivative of

the error function w.

Although there are many rules to optimize the neural

networks developed, the network architecture was often

derived from trial and error approach. Another factor

affecting the convergence of back propagation algorithm

is the learning rate,

. With the large value of

, the

network will increase the learning rate, but the network

with the too large value will not be able to converge.

Inversely, small values can ensure the convergence algo-

rithm, but the learning rate is very slow. For this reason,

the algorithm with adaptive learning rate is applied and

described as follows:

ηηη

∆+=+ )()1( kk

(8a)

and error of learning rate











−=∆

)(

(8b)

where a is the increase coefficient, b is the decrease

coefficient,

)(k

is the kth learning rate.

2.5. Mean Threshold Algorithm

In this project, a threshold algorithm will be applied to

determine cases of open eye, two eyes blinking, glanced

left, glanced right. The average value M of open eye sig-

nals is calculated using the following equation:

ny(n)

∑

(9)

where y(n) is the set of EEG signals (rejected voltage

drift) and N denotes the number of samples.

From Equation (9), the standard deviation SD in case

of open eye signal can be calculated as follows:

nMny

∑

=−

=1))((

(10)

A mean threshold algorithm ThM is built to determine

cases of eyes open, eyes blinks, the left and right glance:

SDa*−= MThM

(11)

where a is the coefficient of the standard deviation.

This paper shows the detection of eye states based on

the change of amplitude of signals with its frequency

range (0.5 to 3.5 Hz) at Fp1, F7 and F8 positions. There-

fore the mean threshold determined based on EEG sig-

nals in the open eye case plays an important role.

To reduce the error of eye blinking recognition, the

threshold value ThM was calculated in the case of open

eye by comparing with the maximum values in the

measured times of eye movements to determine the coef-

ficient a. Therefore, the mean thresholds ThM at the po-

sitions Fp1, F7 and F 8, w ere calculated as follows:

Fp1

OFp1 BFp1

(OFp1, R F p1, LF p1)Max Max

ThM Max

(12)

( )

RBF7OF7 F7

Min <<Min

=Min OF7,LBF7,BF7

ThM (13)

( )

RBF8OF8 F8

Min <<Min

=Min OF8,LBF8,BF8

ThM (14)

where Max is the maximum amplitude at Fp1 of the eye

opening signal (OFp1), right glance signal (RFp1), left

glance signal (LFp1) and MaxBFp1 denotes the maximum

amplitude in the case of blinking eye at Fp1.

3. Results and Discussion

EEG signals were collected at three channels FP1, F7

and F8, in which each channel is processed to extract

features. Mainly two methods of the ARNN model and

the mean threshold algorithm were applied to find the

best method for the wheelchair control.

3.1. Features of Eye Movements Using AR Mode

From the EEG signals of eye movements, each channel

has two AR coefficients. Hence one of the eye states

created a vector with six AR coefficients as shown in

Table 2. Time for an eye activity is less than 1 second,

so the eye activity is just set 1 second. The vectors are

inputs of the feedforward neural networks.

Figures 7(a), 7(b), 8(a) and 8(b) showed the AR

model coefficients of the signals corresponding to eye

open time at 1 second. From the figures, we see that in

the case of eye open and eye blinks, the signals are the

same, so the coefficients of three channels are nearly

equal. While Figures 9(a), 9(b), 10(a) and 10(b) re-

present different coefficients of glancing left and right.

All different coefficients generate four vectors (in Table

2).

Figure 11 represents six coefficients versus the am-

plitude of signals. The c la ssification of four coefficient

vectors shows that they are the same shape of signals but

Table 2. AR coefficient vectors.

Eye activity Coefficient vectors

Opening eyes −1.101 0.203 −1.059 0.189 −1.158 0.186

Blinking eyes −1.952 0.956 −1.715 0.720 −1.566 0.569

Glancing left −1.252 0.396 −1.205 0.266 −1.710 0.892

Glancing right −1.078 0.109 −1.369 0.372 −1.159 0.162

N. T. HAI ET AL.

288

(a)

(b)

Figure 7. (a) Signals of opening eyes; (b) AR model coeffi-

cients.

(a)

(b)

Figure 8. (a) Signals of blinking eyes; (b) AR model coeffi-

cients.

(a)

(b)

Figure 9. (a) Signals of glancing left; (b) AR model coeffi-

cients.

(a)

(b)

Figure 10. (a) Signals of glancing right; (b) AR model coef-

ficients.

00.5 11.5 22.5 33.5 44.5 5

-200

200 Si gnal channel F p1 (F il tered)

Ti me (s)

Ampli tude

00.5 11.5 22.5 33.5 44.5 5

-200

200 Si gnal channel F 7 (F il tered)

Ti me (s)

Ampli tude

00.5 11.5 22.5 33.5 44.5 5

-200

200 Si gnal channel F 8 (F il tered)

Ti me (s)

Ampli tude

123

-1.2

-1

-0.8

-0.6

-0.4

-0.2

0.2

0.4

six cofficients

Ampli tude

00.5 11.5 22.5 33.5 44.5 5

-500

500 Si gnal channel Fp1 (F i l t ered)

Ti me ( s)

Ampli tude

00.5 11.5 22.5 33.5 44.5 5

-200

200 Si gnal channel F7 (F i l t ered)

Ti me ( s)

Ampli tude

00.5 11.5 22.5 33.5 44.5 5

-200

200 Si gnal channel F8 (F i l t ered)

Ti me ( s)

Ampli tude

123456

-2

-1.5

-1

-0.5

0.5

six cofficients

Amplitude

00.5 11.5 22.5 33.5 44.5 5

-500

500 Signal channel F p1 (Fi l tered)

Ti me ( s)

Ampli tude

00.5 11.5 22.5 33.5 44.5 5

-200

200 Signal channel F 7 (Fi l tered)

Ti me ( s)

Ampli tude

00.5 11.5 22.5 33.5 44.5 5

-200

200 Signal channel F 8 (Fi l tered)

Ti me ( s)

Ampli tude

123456

-2

-1.5

-1

-0.5

0.5

six cofficients

Amplit ude

00.5 11.5 22.5 33.5 44.5 5

-500

500 Si gnal channel Fp1 (F i l t ered)

Ti me ( s)

Amplitude

00.5 11.5 22.5 33.5 44.5 5

-200

200 Si gnal channel F7 (F i l t ered)

Ti me ( s)

Amplitude

00.5 11.5 22.5 33.5 44.5 5

-200

200 Si gnal channel F8 (F i l t ered)

Ti me ( s)

Amplitude

1 2 3 4 5 6

-1. 4

-1. 2

-1

-0. 8

-0. 6

-0. 4

-0. 2

0.2

0.4

six cofficients

Ampli tude

N. T. HAI ET AL.

289

Figure 11. Representation of six coefficients of the eye ac-

tivities.

their amplitudes are different. These coefficient vectors

will be applied to inputs of the neural network model for

recognizing the eye activities.

3.2. Identification o f E ye Movements Using NN

Subjects worked out their tasks for recordings, in which

20 times blink eyes, 15 times glanced left, glanced right

15 times and 20 times to open eyes. Thus, we have a total

of 70 vector samples, in which 50 sample vectors (eye

blinking_15, glancing le ft_10, glancing right_10 and

opening eyes_10) will be used to train the Artificial

Neural Networks (ANNs) and the 20 remaining vectors

will be applied to check the training results.

The experiment using a two hidden layers network

structure with sigmoid function has its output is a linear

function as described in Table 3. In this experiment, the

number of hidden layer neurons was chosen as in Table

4, in which the first hidden layer with 15 neurons and the

number of neurons in the second hidden layer will be

changed for investigating the accuracy of the network.

From Table 4, we see that the network with the 25 neu-

rons in the 2nd hidden layer is the highest with the aver-

age accuracy of 94%.

In this paper, the feed forward neural networks with

back-propagation learning rule using gradient reduction

algorithm were used, in which learning rate is 0.001

(learning rate is the smaller, training time is longer, but

the obtained results are more accurate). While the num-

ber of iterations is 1000 (the number of iterations is as

large as possible, because the error of the network out-

puts and the real outputs is smaller), the increasing ratio

of learning rate is a = 1.07 and the decreasing ratio of

learning rate b = 0.7.

3.3. Eye Movements Using M ean Threshold

From Equations (11), (12) and (13), one determined the

coefficients a, in which 3.5 < aFp1 < 13.75 and we chose

aFp1 = 11 at the Fp1 position for the case of eye blink s and

similarly, aF7 = −4 and aF8 = −4 were chosen for the posi-

tions, F7 and F8. Based on the v alues aFp1 = 11, aF7 = −4

and aF8 = −4, the mean thresholds were calculated as

Table 3. Description of the NN outputs.

Eye activity Desired outputs

Opening eyes 1 0 0 0

Blinking eyes 0 1 0 0

Glancing left 0 0 1 0

Glancing right 0 0 0 1

Table 4. Results using neural networks.

Number of

hidden layer

neurons

Accuracy (%)

Eyes

open Eyes

Blinks Left

Glance Right

Glance Average

15 * 10 95 95 78 85 89

15 * 20 95 90 90 85 90

15 * 25 90 97 92 95 94

15 * 30 90 95 90 90 92

shown in Table 5 In similarity, the mean threshold val-

ues were obtained on nine subjects as shown in Table 6.

The mean threshold values were applied into eye tasks

and we recognized the eye activities times (see Figure

12).

From Table 7, the ANN method gives the higher per-

formance. However, time for training data is more ex-

pensive. Therefore, which method chosen here is depen-

dent on each typical application.

The author et al. applied the SVMs and ANNs for the

eye movements using EEG and showed the classification

accuracies, in which the SVMs is 90.8% and accuracy of

86.8% is of the ANNs [13]. While based on AR coeffi-

cients, the ANNs have accuracy of 93.5% and the accu-

racy of the mean threshold is 86.25% in this paper. This

means that our proposed methods are the effectiveness.

3.4. Wheelchair Control Strategy

In a Brain-Computer Interface (BCI) system for control

of an electric wheelchair as shown in Figure 13, the user

was concentrating to drive the wheelchair by the eye

movements. Figure 14 shows the directions of the elec-

tric wheelchair, in which the wheelchair can be driven to

move with commands such as forward, backward, stop,

turning left and turning right.

The wheelchair was designed to move with the speed

of 5 km/h in the indoor environment. For the smooth

movement of the wheelchair, when the wheelchair rece-

ives a typical command to move to the left or the right, it

was designed to follow a curve around the inflection

point of the cubic equa tion.

4. Conclusion

This paper investigated an AR neural network algorithm

and the mean threshold algorithm in an EEG-controlled

-2 .5

-2

-1 .5

-1

-0 .5

0 .5

1 .5

012

34567

Coefficient

Amplitude

Opening eyes

Blinking eyes

Glancing left

Glancing right

N. T. HAI ET AL.

290

Table 5. Experimental results.

Times

Fp1 F7 F8

Eye Blink Eye open Right glance Eye open Left glance Eye open

MaxFp1 MaxBFp1 ThMOFp1 MinRB F7 MaxOF7 ThMOF7 MaxRBF8 MinOF8 ThMOF8

1 56 166 121 −91 −32 −120 −112 −45 −56

2 30 182 154 −103 −33 −128 −146 −24 −64

3 39 171 154 −112 −38 −44 −89 −35 −84

4 29 164 143 −133 −61 −80 −125 −39 −108

5 58 147 110 −110 −59 −76 −117 −33 −112

6 29 163 110 −112 −22 −76 −109 −25 −112

Mean 40 165.5 132 −110 −40 −84 −116 −33.5 −89

Table 6. The mean thresholds.

Subject ThMOF1 ThMOF7 ThMOF8

S1 132 −84 −89

S2 134 −100 −93

S3 145 −89 −87

S4 156 −80 −80

S5 160 −112 −100

S6 142 −102 −87

S7 115 −98 −85

S8 143 −87 −80

S9 165 −90 −98

Mean 143 −93 −88

Table 7. Results of two methods.

Method

Accuracy (%)

Eyes

open Eyes

Blink Left

Glance Right

Glance Average

ANN 90 97 92 95 93.5

ThM 85 90 85 85 86.25

wheelchair for severely disabled people. From original

signals, the Hamming low pass filter was applied to pro-

duce the frequency bands for feature extraction. The

coefficients, which bring the feature of each eye activity,

are extracted using the AR model. These coefficients

generated the feature vectors for connecting to inputs of

the neural network which was employed to recognize the

eye movements such as opening eyes, blinking eyes,

glancing left and glancing right. After recognizing these

characteristics, user can drive the wheelchair to reach the

target. Experimental results showed that the wheelchair

user can move in the indoor environment.

(a)

(b)

(c)

Figure 12. (a) The threshold ThMOFp1; (b) The threshold

ThMOF7; (c) The threshol d ThMOF8.

00.5 11.5 22.5 33.5 44.5 5

-100

-50

100

150

200

250 Si gnal c hannel F p1 (F i ltered)

Ti me ( s)

Ampli tude

THR

OFp1

=143

Eyes open

Eyes blinks

00.5 11.5 22.5 33.5 44.5 5

-140

-120

-100

-80

-60

-40

-20

60 Si gnal channel F 7 (Fi l tered)

Ti me (s)

Amplitude

THR

OF7

=-93

Eyes open

Glanced right

00.5 11.5 22.5 33.5 44.5 5

-120

-100

-80

-60

-40

-20

40 Si gnal channel F 8 (Fi l tered)

Ti me ( s)

Amplitude

THR

OF8

=-88

Eyes open

Glanced left

N. T. HAI ET AL.

291

Figure 13. A user is controlling the wheelchair.

Figure 14. The directions of the wheelchair motion.

5. Acknowledgements

We would like to thank Vietnam National University in

Ho Chi Minh City for supporting research grant No.

C2013-28-06. Furthermore research was partly supported

by a research fund from International University in Ho

Chi Minh City. Finally, an honorable mention goes to our

volunteers and colleagues for supports on us in complet-

ing this project.

REFERENCES

[1] J. Wolpaw, N. Birbaumer, D. McFarland, G. Pfurtschel-

lere and T. Vaughan, “Brain-Computer Interfaces for

Communication and Control,” Clinical Neurophysiology,

2002, pp. 767-791.

http://dx.doi.org/10.1016/S1388-2457(02)00057-3

[2] N. Ince, F. Goksu, A. Tewfik and S. Arica, “Adapting

Subject Specific Motor Imagery EEG Patterns in Space-

Time-Frequency for a Brain Computer Interface,” Bio-

medical Signal Processing and Control, 2009, pp. 236-

246. http://dx.doi.org/10.1016/j.bspc.2009.03.005

[3] N. Weiskopf, F. Scharnowski, R. Veit, R. Goebel, N.

Birbaumer and K. Mathiak, “Self-Regulation of Local

Brain Activity Using Real-Time Functional Magnetic

Resonance Imaging (fMRI),” Journal of Physiology,

2004, pp. 357-373.

[4] S. Lloyd-Fox, A. Blasi and ElwellCE, “Illuminating the

Developing Brain: The Past, Present and Future of Func-

tional Near Infrared Spectroscopy,” Neuroscience-Bio-

behavioral, 2010, pp. 269-284.

[5] G. E. Fabiani, D. J. McFarland, J. R. Wolpaw and G.

Pfurtscheller, “Conversion of EEG Activity into Cursor

Movement by a Brain-Computer Interface,” Transactions

on Neural Systems and Rehabilitation Engineering, Vol.

12, pp. 331-338.

http://dx.doi.org/10.1109/TNSRE.2004.834627

[6] C. Guger, W. Harkam, C. Hertnaes and G. Pfurtscheller,

“Prosthetic Control by an EEG-Based Brain-Computer

Interface (BCI),” IEEE Transactions on Robotics, Vol. 21,

2005.

[7] D. J. Krusienski and J. J. Shih, “A Case Study on the

Relation between Electroencephalographic and Electro-

corticographic Event-Related Potentials,” 32nd Annual

International Conference of the IEEE EMBS, 2010.

[8] D. J. McFarland and J. R. Wolpaw, “EEG-Based Com-

munication and Control: Speed-Accuracy Relationships,”

Applied Psychophysiology and Biofeedback, Vol. 28,

2003, pp. 217-231.

http://dx.doi.org/10.1023/A:1024685214655

[9] B. Blankertz, G. Dornhege, M. Krauledat, K.-R. Müller,

V. Kunzmann, F. Losch and G. Curio, “The Berlin Brain-

Computer Interface: EEG-Based Communication Without

Subject Training,” IEEE Transactions on Neural Systems

and Rehabilitation Engineering, Vol. 14, 2006, pp. 147-

152. http://dx.doi.org/10.1109/TNSRE.2006.875557

[10] X. Gao, D. Xu, M. Cheng and S. Gao, “A BCI-Based

Environmental Controller for the Motion-Disabled,” IEEE

Transactions on Neural Systems and Rehabilitation En-

gineering, Vol. 11, 2003, pp. 137-140.

http://dx.doi.org/10.1109/TNSRE.2003.814449

[11] K. S. Ahmed, “Wheelchair Movement Control VIA Hu-

man Eye Blinks,” American Journal of Biomedical Engi-

neering, Vol. 1, 2011, pp. 55-58.

[12] T. Q. D. Khoa and M. Nakagawa, “Functional Near

Infrared Spectroscope for Cognition Brain Tasks by

Wavelets Analysis and Neural Networks,” International

Journal of Biological and Life Sciences, Vol. 4, 2008, pp.

28-33.

[13] R. Singla, B. Chambayil, A. Khosla and J. Santosh,

“Comparison of SVM and ANN for Classification of Eye

Events in EEG,” Journal of Biomedical Science and En-

gineering, Vol. 4, 2011, pp. 62-69.

[14] S. Chabaa, A. Zeroual and J. Antari, “Identification and

Prediction of Internet Traffic Using Artificial Neural

Networks,” Intelligent Systems & Applications, Vol. 2,

2010, pp. 147-155.

http://dx.doi.org/10.4236/jilsa.2010.23018

[15] N.-J. Huan and R. Palaniappan, “Neural Network Clas-

siﬁcation of Autoregressive Features from Electroence-

phalogram Signals for Brain-Computer Interface Design,”

Journal of Neural Engineering, Vol. 1, 2004, pp. 142-150.

http://dx.doi.org/10.1088/1741-2560/1/3/003

Wheelchair

Computer

Active Two

EEG

Electrod e

For wa r d

Backward

Turn right

Turn left