Proceedings of the Eighth International Conference on Machine Learning and Cybernetics, Baoding, 12-15 July 2009

978-1-4244-3703-0/09/$25.00 2009 IEEE

3342
EOG-BASED SIGNAL DETECTION AND VERIFICATION FOR HCI
LAWRENCE Y. DENG
1
, CHUN-LIANG HSU
2
, TZU-CHING LIN
1
, JUI-SEN TUAN
1
, YUNG-HUI CHEN
3

1
Department of Computer Science and Information Engineering, St. Johns University
2
Department of Electrical Engineering, St. Johns University
3
Department of Computer Information and Network Engineering, Lunghwa University of Science and Technology
E-MAIL: [email protected]
Abstract:
In this paper, we proposed an eye-movement tracking
system. Based on Electro-Oculography (E.O.G) technology we
detected the signal with different directions in eye-movements
and then analyzed to understand what they represented about
(e.g. horizontal direction or vertical direction). We converted
the analog signal to digital signal and then used as the control
signals for Human Computer Interface (HCI). In order to
make the system robust, several applications with
EOG-based HCI had been designed. Our preliminary results
revealed more than 90% accuracy rate for examining the
eye-movement that may become a new useful human-machine
user interface in the near future.
Keywords:
Electro-Oculography (E.O.G), Motor Neuron Disease,
Eye-Movement, Amyotrophic Lateral Sclerosis,
Human-Machine/Computer Interface (HMI/HCI)
1. Introduction
Jean-Dominique Bauby was a famous journalist and
editor of the fashion magazine ELLE. In 1995, Bauby
suffered a massive stroke. His condition is called Locked-in
Syndrome; he could only blink his left eyelid. In spite of his
hard condition, Bauby blink his eye when the correct letter
was reached by his interlocutor reciting the alphabet and
finally accomplished the book The Diving Bell and the
Butterfly in 1997 [9]. In Taiwan, Mr. Chen suffered Motor
Neuron Disease, (M.N.D) or Lou Gehrig's disease.
Impressively, Mr. Chen accomplished 5 books that contain
190,185 words from 1999 to 2005 (the most words written
by blinking in the Guinness World Records) [6]. For
thousands of people, Amyotrophic Lateral Sclerosis (ALS)
deprives them from using their limbs or even speech. They
leave the brain and eyes activity unimpaired. The only way
for communication is to detect the Eye-movement
correctness. EOG-based signal processes provided the
detection of the Eye-movement resulting from looking-up,
down, left and right. [13] have developed a new method for
automatic estimation of vigilance level by using
electroencephalogram (EEG), electromyogram (EMG) and
eye movement (EOG). [3,4] diagnosis of subnormal eye
through the analysis of EOG signals and then to assist the
physician to make the final decision without hesitation. So
the eye-movement HCI can be taken as an input device for
communication [8,11,16,18].


Figure 1. Eye-movement HCI platform architecture.
In this paper, we constructed an eye-movement HCI
platform for several applications. Figure 1 illustrated the
Eye-movement HCI platform architecture. The
eye-movement coils are the electromagnetic fields that
received signal between 50 to 3500uV with a frequency
range of about dc-100 Hz. Eye-movement signals needs to
be adjusted/normalized by signal amplifier, adders and
filters. The filters eliminate the effects due to other
bio-potentials, such as saccadic/blink over to the EOG
signal. After that, the EOG signal is then converted to
Universal Asynchronous
Receiver/Transmitter, UART
Eyemovement coils
Signal Amplifier, adder and
Analog-to-Digital
Computer COM port (RS232),
Applications Control
AP 1 AP 2 AP n
Proceedings of the Eighth International Conference on Machine Learning and Cybernetics, Baoding, 12-15 July 2009
3343
digital signal and then transmitted to the computer. The
application control component will deal with the
eye-movement and then interact with the desired
application. Therefore, our proposed platform can serve
various services.
The paper is organized as follows. Section 2
introduces the related works with Eye-movement detection.
Some classical eye-movement measurement techniques for
recording the electrical activity of eye-movement will be
described in details in this section. The proposed
EOG-based Eye-movement HCI platform is given in
section 3. Some propose prototype applications for
Eye-movement are shown in section 4. The diverse
guidance tests and experimental results are given in section
5. Finally, section 6 draws the conclusion and future work.
2. Eye-movement Measurement Techniques
Some electrodes measurement for bio-potential
recordings and suitable applications are discussed in this
section. These related eye-movement measurement
techniques and applications contains: Eye-Mouse (EM)
[8,14], Eye-writing System and emulation of Oral Speaking,
Eye-Tracking System for powered wheelchair [17]and
EOG-based application for Hospital Alarm System [16].
a. Infrared Video System, IRVS
Infrared Video System (IRVS) consists of a
charge-coupled device (CCD) camera, an image capture
card, and an LCD projector [7,19]. These methods are
suitable for real-time extraction of the position of the pupil
in the eye image. Since the eye gaze tracking may represent
a persons focus of attention, it has potential applications in
Human Computer Interaction (HCI), Virtual Reality,
Eye-Disease Diagnosis, and Human Behavior Studies. In
addition, some of these systems required the users head to
be motionless during eye tracking. They must often be
calibrated repeatedly for each individual, and they also
showed low tolerance for head movements and required the
user to keep their heads still.
b. Infrared Oculography, IROG
In the infrared Oculography (IROG), the infrared light
reflected through Sclera, is converted into current to detect
the movement of eyes [12]. It is cheaper and can be easily
applied. However, the harm caused by infrared light to eyes
is unknown.
c. Search Coil (SC)
Search Coil (SC) [15], which was inducted by
magnetic field, possessed benefits as high precision and
fixed magnetic field. But it couldnt use too long due to the
influence of excretions from eyes.
d. Optical-type Eye Tracking System
Optical-type Eye Tracking System was composed of
headset displayer, mini-CCD, light source, image-fetching
card, and image processing unit. The method fetches the
images of eyes with a mini-CCD and then figures out the
central position of pupil with a specific program [1,17]. Its
benefit included noncontact, and could be used after putting
on the headset. But the weight of the headset and bright
light source would make users uncomfortable.
e. Purkinje Dual-Purkinje-Image (DPI)
Purkinje DPI is to analysis four Purkinje images
reflected from the incoming light on the boundaries of the
lens and cornea. Tracking the direction of gaze by the pupil
tracking technique is similar to limbus tracking, but only
the smaller boundary between the pupil and the iris is used
here [2]. DPI method provided a more tracking accuracy
because it avoids the shelter issues from the eyelid. Anyway,
the related set of system is more expensive.
f. Electro-Oculography (EOG)
Electro-Oculography (EOG) detects the eye movement
by recording the corneal-retinal potential difference from
hyper-polarizations and de-polarizations existing between
the cornea and the retina. The potential difference can be
taken as the electrical dipole with a negative pole at the
fundus and positive pole at the cornea. By measuring the
voltage induced across a set of electrode coils pasted
around the eyes as, we can measure the electric signal of
the eyes dipole as the eye-movement changes [8,11,16,18].
EOG is the most widely used technique in measuring the
bio-potential.
3. Eye-movement HCI System
a. System Architecture
We designed the hardware circuits, and according to
the literature [14], the eye-movement coils produces
between 50 to 3500uV with a frequency ranging from
dc-100 Hz. So we must design the amplified circuits. The
flow chart of detecting signals is shown in Fig.2. The EOG
signal is input to the instrumentation amplifier, in which
the signals are amplified by about 50 times. The circuit uses
an adder to balance the signals, 0.05 Hz high pass filter and
another amplifier with a gain of 10 to adjust the signals to
1.5V. The flow chart of EOG signal recordings is illustrated
in figure 2.

Proceedings of the Eighth International Conference on Machine Learning and Cybernetics, Baoding, 12-15 July 2009
3344

Figure 2. Flow chart of the EOG signal recording.
A.1 Bio-potential Recording
The EOG was attached on to the surrounding area of
eyes with the help of others, and then turned on to detect
the corneal-retinal potential difference. We chose pragmatic
reliable low-cost electrodes for long-term use in EOG
signal recordings.
A.2 Bio-potential Amplifier
The amplifier IC (INA128, Texas Instruments, TI) as
shown in figure 3, which was composed of 3 Difference
amplifiers was used. We have to modify the RG resistance
to change its gain [17]. Here we adapted as to
get a voltage gain of about 50 times.
Vo=
) )(
2 2
1 (
3
4
+
+ V V
RG
R
R
R







Figure 3. INA128 [17]
A.3 Balance
Signals differentially amplified by INA128 may not be
balanced, so we adapted an adder to balance the signal.
A.4 Filtering
According to literature, the signal band-width of
eye-movement was located in the range of 0.1Hz~20Hz, so
signals more than 20Hz would be regarded as noise which
should be deleted using noise filters. Here we designed a
band-pass filter [14] ranging from 00.5 to 23 Hz to isolate
the disturbance from grid electricity of 60Hz. However, we
realized that just a one-class low-pass filter couldnt
effectively avoid the noise, so we added another low-pass
filter of 19Hz, and the output was better with much better
stability.

fH=
2
2 1 2 1
1
C C R R

=
) 10 1 10 3 ( 2
1
6 6


=
3 28 . 6
1

=
84 . 18
1
=0.053 Hz

Figure 4. High pass filter.

The high pass filter is shown in figure 4, in which the
signal frequency below than 0.05 Hz would be isolated. In
figure 5, the first-class low-pass filter is used to isolate
signals of frequency higher than 23.417Hz. In Fig 6, the
second-class low-pass filter would deal with the signals
passing through the first-class filter to get a higher Q value
of signals.
fH=
2
4 3 10 9
1
C C R R

=
) 10 1 . 0 10 68 ( 2
1
6 3


=
3
10 8 . 6 28 . 6
1

=
3
10 704 . 42
1

=23.417Hz


Figure 5. First-class low-pass filter
fH=
2
6 5 14 13
1
C C R R

=
) 10 1 . 0 10 82 ( 2
1
6 3


=
3
10 2 . 8 28 . 6
1

=
3
10 496 . 51
1

=19.418Hz

Figure 6. Second-class low-pass filter.
A.5 Horizontal Movement
Since the voltage range of AD converter in our
microprocessor was from 0 to 3.3V, we needed to design an
Voltage output to SPCE061A
Instrument amplifier
Adder
High-pass filter of 0.05Hz
Voltage adjusting and filting
Low-pass filter of 23Hz
Low-pass filter of 19Hz
Sinnals input of EOG
Proceedings of the Eighth International Conference on Machine Learning and Cybernetics, Baoding, 12-15 July 2009
3345
adjusting circuit to let the detected voltage available.
A.6 Input into microprocessor
The reason why we selected SPCE061A as our core
microprocessor was due to its advantages like
re-programming capability, real-time debugging, low cost,
and simple interface. After AD converting, the signals are
sent to the microprocessor through an RS-232 interface
[10].
B. Preliminary Experimental Results
The following paragraph will illustrate the EOG
analog signal for bio-potential changes when the eye moves
up, down, left, and right.
B.1 EOG Analog Signal
Figure 7 show the fuzzy set values for the EOG analog
signal waves recorded by the Oscillograph. The related
fuzzy distinction rules are shown in Table 1. More testing
results are shown in Appendix I.
Table 1: Fuzzy Distinction Rule
If A is high and B is high, then Up.
If A is mid and B is mid, then Right.
If A is mid and B is mid, then Stop.
If A is mid and B is low, then Down.
If A is low and B is mid, then Left.
Where A is Left-Right signal and B is Up-Down
signal.


Figure 7. Fuzzy Set values for EOG
B.2 EOG Digital Signal Detection
The signals after A/D conversion are sent to a
computer through the RS232 interface. The signal detection
flow chart is shown in Figure 8.
timer1_Tick {
If(isTestOk == true) {
if(num==stopsignal){
stopcount++;
if (stopcount > 20)
waitFlag = true;
} }
else stopcount = 0
if(waitFlag == true){
if(num==Up Dn Lf Rt signal)
{
dirction=1,2,3,4;
waitFlag = false;
stopcount = 0;
}}

Figure 8. RS232 signal detection flow chart.
4. Detection of Eye-Movement
After designing three kinds of detecting the interface
of eye-movement, the system could be applied into the
system of controlling TV, DIY eye-sight level checking, and
eye-controlling game.
A Platform of control with eye-movement
Application selection user interface is shown in Figure
9.
if (rs232.isStart== true){
if (Program.rs.direction==direction){
textBox3.Text=direction;
switch (Up Dn Lf Rt signal){
3 case
break;
}
if (direction count >= 3) {
direction count = 0;
timer1.Enabled = false;
rs232.timerFlag = false;
Show();
}
} }
Ti
pe
riod0.1
Sto
p?
Stop
counter=0
Stop
counter+1
Stop
counter>=20
u d l ri
U Dow Le Rig
Stop
Proceedings of the Eighth International Conference on Machine Learning and Cybernetics, Baoding, 12-15 July 2009
3346

Figure 9. Applications Selection view.

B. The adjusting system of eye-movement
As shown in figure 10, whenever the users open the
system, the system would execute the adjusting system
automatically.
The design could meet with the eye-movement ranges
of different users.
The system could record and figure out the different
movement ranges of users to judge the real action of eyes.
After the adjustment of the system, the precision of the
whole system could get up to 90%.
need help with others.
button1_Click
if (timer1.Enabled){
timer1.Enabled = false;
}
else{
timer1.Enabled = true;
}
button8_Click
{
if (!File.Exists("record.txt"))
Sw=File.CreateText("record.txt");
else
sw=File.AppendText("record.txt");
}

Figure 10. EOG signal adjustment sub-system view.

C. EOG based remote-control System for TV
Turn on the TV with Electro-Oculography interface as
shown in Fig 11. Its function is to shift the channels and to
adjust the volume. timer1.Enabled = true;
pInfo.UseShellExecute = true;
if (rs232.isStart==true){
if (P.r.d==direction){
textBox1.Text = direction;
if (UpDnLfRtsignal==direction){
rs232.count = 0;
SendKeys.Flush();
}
}
}


Figure 11. EOG based Remote-control System for TV view.
D. Eye-sight Level Checking System
Upon starting the system, it would give out the test
items, and users could answer the test items with
Electro-Oculography interface and test out the eyesight
level as shown in Fig.12.
After comparing with the level tested by the hospital, a
precision of up to 95% could be obtained.
It worked without any help from others.
if (rs232.isStart == true){
label2.Text
if (Up Dn Lf Rt signal ==direction){
if (UpDnLfRtsignal ==direction)
r++;
else
w++;
nextPic();
count = 0;
}
}
if (end == true && final == true && r == 2){
end = false;
}
Proceedings of the Eighth International Conference on Machine Learning and Cybernetics, Baoding, 12-15 July 2009
3347

Figure 12. EOG-based eye-sight level checking system view
E. EOG-based Game with Eye-control
The practical game interface is shown in Figure 13.
new PictureBox[15];
for(int i=0;i<3;i++)
for(int j=0;j<5;j++){
picrecord[i * 5 + j] = index;
}
Up{
if (picrecord[5 * i + mypicloc] == mypicindex){
pic[5*i+mypicloc].Image=null;
picrecord[5*i+mypicloc]=99;
if (count == 15){
Show("Win!!!")
break;}
}
}
Down{
mypicindex=(++mypicindex)%3;
}
Left or right{
mypicloc--;(mypicloc++;)
}

Figure 13. EOG-based Game with eye-control view.
5. Experimental Discussion
A. Distortion of EOG Signals
In the experimental processes, we found that there
would be 30 to 90 minutes of best signal stability for every
user with different physical conditions and environment,
and over the time period, the signals would shift. We tried
to explain this situation as resulting from long time
attachment of EOG and sweating of skin. If we wash the
EOG with alcohol, the situation would be improved.
A.2 Movement of Eyes
Since the range of horizontal movement of eyes was
wider than the vertical movement, the detected signals
became significant and larger and did not easily cause
error .
A.3 Face muscle and eyes movement
If the users are nervous, blink eyes, or make much
exaggerated gestures and speech, all of such actions would
cause irregular movement of face muscle and eyes to
decrease the precision of detecting eye-movement.
6. Conclusions
In this paper, we proposed a multi-purposes
eye-movement tracking system. Integrating
Electro-Oculography (E.O.G) installations we detected the
signal with different directions in eye-movements and then
analyzed to understand what they represented about (e.g.
horizontal direction or vertical direction). We converted the
analog signal to digital signal and then used as the control
signals for Human Computer Interface (HCI). We finally
accomplished the multi-purposes eye-movement HCI
platform, and this system cost was much lower than other
methods. The system adapted the direction-sensor interface
design, and the system would be active according to the
indicated direction. Users could easily use this system and
the precision of the system after adjustment could get more
than 90%.
Acknowledgements
This study was awarded the prize of Masterpiece in the
Instrument Group in the Microprocessor Design and
Application Contest in 2008 authorized by the Taiwan
Education Ministry.
References
[1] A. Geetha, V. Ramalingam, S. Palanivel, B.
Palaniappan, Facial expression recognition A real
time approach, Expert Systems with Applications,
36(1), 303-308, 2009
[2] A. J. Glenstrup and T. E. Nielsen, Eye Controlled
Media: Present and Future State, Thesis of Bachelor in
Information Psychology, Psychological Laboratory,
University of Copenhagen, Denmark. 1995
[3] Ayegl Gven, Sadk Kara, Classification of
electro-oculogram signals using artificial neural
Proceedings of the Eighth International Conference on Machine Learning and Cybernetics, Baoding, 12-15 July 2009
3348
network, Expert Systems with Applications, 31(1),
199-205, 2006
[4] Ayegl Gven, Sadk Kara, Diagnosis of the
macular diseases from pattern electroretinography
signals using artificial neural networks, Expert
Systems with Applications, 30(2), 361-366, 2006.
[5] Budi Thomas Jap, Sara Lal, Peter Fischer, Evangelos
Bekiaris, Using EEG spectral components to assess
algorithms for detecting fatigue, Expert Systems with
Applications, 36(2-1), 2352-2359, 2009.
[6] ChungHwa, Digi-image, Alliance Co. Ltd., 2002,
http://www.nownews.com/2002/03/13/91-1274722.ht
m
[7] C. S. Lin, C. C. Huan, C. N. Chan, M. S. Yeh, and C.
C. Chiu, Design of a computer game using an
eye-tracking device for eye's activity rehabilitation,
Optics and Lasers in Engineering, 91-108, 2004.
[8] G. Norris and E. Wilson, The Eye Mouse, An Eye
Communication Device, IEEE Proceedings of
Bioengineering, 6667, 1997.
[9] Jean-Dominique Bauby, The Diving Bell and the
Butterfly, 1997.
http://divingbellandbutterfly.cmcmovie.com/
[10] Sunplus Technology Co., Ltd. http://w3.sunplus.com/
[11] K. S. Park and K. T. Lee, Eye-controlled human
computer interface using the line-of-sight and the
intentional blink, Computer Engineering, 30(3),
463-473, 1996.
[12] Kumar, A., and Krol, G., Binocular infrared
oculography. Laryngoscope, 102, 367378, 1992.
[13] Muhammed B. Kurt, Necmettin Sezgin, Mehmet Akin,
Gokhan Kirbas, Muhittin Bayram, The ANN-based
computing of drowsy level, Expert Systems with
Applications, 36(2-1), 2534-2542, 2009.
[14] Rangaraj M. Rangayyan, Biomedical Signal Analysis:
A Case Study Approach, John Wiley and Sons Inc.,
First Edition, 2001.
[15] R. V. Kenyon, A Soft Contact Lens Search Coil for
Measuring eye Movements, Vision Res. 25(11),
1629-1633, 1985.
[16] S. Venkataramanan, Pranay Prabhat, Shubhodeep Roy
Choudhury, Harshal B. Nemade, and J.S. Sahambi,
Biomedical Instrumentation based on EOG Signal
Processing and Application to a Hospital Alarm
System, Proc. of IEEE ICISIP 2005, 535-540, 2005.
[17] Texas Instruments, http://www.ti.com/, 2008.
[18] Y. Tomita, Y. Igarashi, S. Honda, and N. Matsuo,
Electro-Oculography Mouse for Amyotrophic Lateral
Sclerosis Patients, IEEE Conference Engineering in
Medicine and Biology Society, 531, 1708-1781, 1996.
[19] Z. Zhu and Q. Ji, Eye gaze tracking under natural
head movements, IEEE Computer Society
Conference on Computer Vision and Pattern
Recognition, 1, 918-925, 2005.