Development of a human computer Interface system using EOG

doi:10.4236/health.2009.11008

Paper Menu >>

Journal Menu >>

HEALTH, 2009, 1, 39-46

3939Published Online June 2009 in SciRes. http://www.scirp.org/journal/health

Development of a human computer Interface system

using EOG

Zhao Lv1,2, Xiao-Pei Wu1, Mi Li2, De-Xiang Zhang1

1The Key Laboratory of Intelligent Computing & Signal Processing, Ministry of Education China, of Anhui University, Hefei, Anhui

Province; 2The First Aeronautical College of Air-Force, Xinyang, Henan Province, China.

Email: kjlz@163.com

Received 13 April 2009; revised 1 May 2009; accepted 7 May 2009.

ABSTRACT

Bio-based human computer interface (HCI) has

attracted more and more attention of researches

all over the world in recent years. In this paper, a

HCI system which based on electrooculogram

(EOG) is proposed. It transforms electrical po-

tentials recorded by horizontal and vertical EOG

into a computer in order to control external

equipment. The system consists of EOG acqui-

sition unit, EOG pattern recognition part and

control command output unit. Three plane elec-

trodes are employed to detect EOG signals,

which contain the information related to the eye

blinking and vertical (or horizontal) eye move-

ments referred to pre-designed command table.

An online signal processing algorithm is de-

signed to get the command information con-

tained in EOG signals, and these commands

could be used to control the computer or other

instruments. Based on this HCI system, the

remote control experiments driven by EOG are

realized.

Keywor ds: human computer interface (HCI); elec-

trooculogram (EOG); blinking; eye movements

1. INTRODUCTION

Bio-based human computer interface (HCI) has the po-

tential to enable severely disabled people to drive com-

puters directly by bioelectricity rather than by physical

means. A study on the group of persons with severe dis-

abilities shows that many of them have the ability to con-

trol their eye movements, which could be used to develop

new human computer interface systems to help them

communicate with other persons or control some special

instruments. Furthermore, this application of EOG-based

HCI could be extended to the group of normal persons

for game or other entertainments. Nowadays, some

methods which attain user’s eye movements are devel-

oped. For example, eye tracking is a technology in which

a camera or imaging system visually tracks some features

of the eye and then a computer determines where the user

is looking at. Eye tracking technology can be divided into

two areas; firstly a remote computer-mounted device, in

which an IR camera is mounted on a computer screen,

and secondly a head-mounted device, in which an IR

camera is placed on user’s head, so this method can get

the eye’s position accurately. However, this technique is

too expensive.

Electro-oculography (EOG) is a new technology of

placing electrodes on user’s forehead around the eyes to

record eye movements. EOG is a very small electrical

potential that can be detected using electrodes. Compared

with the EEG, EOG signals have the characteristics as

follows: the amplitude is relatively high (15-200uV ), the

relationship between EOG and eye movements is linear,

and the waveform is easy to detect. Considering the

characteristics of EOG mentioned above, EOG based

HCI is becoming the hotspot of bio-based HCI research

in recent years. In this paper we present an on-line EOG

signals acquisition and detection system based on EOG

pulses, which are generated by a group of eye movements,

such as looking up and down, looking right and left,

blinking two, three or four times, etc. This system in-

cludes the EOG acquisition circuit, the EOG processing

and commands output part. In order to get the number of

blinking, the system firstly suppress some interference

which exists in the EOG signal, and then remove the

mean value and use dynamic threshold to normalize them,

furthermore, these normalized pulsed is derivation to a

negative and a positive impulse, we neglect negative im-

pulse and count positive impulse, in this way the number

of blinking is attained. On the other hand, we design a set

of “Referenced Pulses” and multiply them by “Normal-

ized EOG signals” to get some new information in order

to detect the eye movements.

The paper is organized as follows: Section 2 intro-

duces the fundamental principles of the propose HCI

system based on EOG. Section 3 describes the imple-

mentation of the system in detail. In this section, firstly it

40 Z. Lv et al. / HEALTH 1 (2009) 39-46

depicts how to acquire EOG signals and analysis the de-

sign circuit. Next, the course of EOG processing, which

includes designing the 50Hz narrow notch filter, using

dynamic threshold to normalize the EOG signals, is intro-

duced. Lastly, it analyses some algorithms of detecting of

the blinking and eye movements. The experiments are

shown in Section 4 and the conclusion is given in Section 5.

2. Base Fundamental Principles of

System

2.1. EOG Detection

The electrooculogram (EOG) is the electrical signal pro-

duced by the potential difference between the retina and

the cornea of the eye [1]. This difference is due to the

large presence of electrically active nerves in the retina

compared to the front of the eye. Many experiments show

that the corneal part is a positive pole and the retina part

is a negative pole in the eyeball. Eye movement will re-

spectively generates voltage up to 16uV and 14uV per

1° in horizontal and vertical way [2]. The typical EOG

waveforms generated by eye movements are shown in

Figure 1.

In Figure 1, positive or negative pulses will be generated

when the eyes rolling upward or downward. The amplitude

of pulse will be increased with the increment of rolling an-

gle, and the width of the positive (negative) pulse is propor-

tional to the duration of the eyeball rolling process.

—

＋

Retina

Corneal

—

＋

Retina

Corneal

③ rolling eyes downward

— ＋

Corneal

①

look strai

ht ahead ② rolling eyes upward

Retina

Figure 1. Eye movements and the corresponding waveform.

In our HCI system, three electrodes are employed to at-

tain the EOG signals. Figure 2 shows the electrode

placement. The horizontal plane electrode (①) is posi-

tioned on the temples to acquire horizontal EOG signal,

and vertical electrode (②) is placed roughly above the

midline of the eye to get the vertical EOG and eyze

blinking signals. The reference electrode is placed at the

mastoid.

2.2. Basic Components of the System

The proposed HCI system block diagram is shown in

Figure 3, which is composed of three parts: EOG acqui-

②

①

Figure 2. Positions of electrodes.

EOG

Acquisition

Circuit

USB

Figure 3. Basic block component diagram of HCI system.

Blinking

Detection

Noise Removal

Eye Movements

Detection

Pre-processing

Mean value Removal

Normalization

Processing

Output Control

Commands

Recognition results

Display EOG Save EOG

signals

waveforms

Z. Lv et al./ HEALTH 1 (2009) 39-46 41

sition module, EOG signals recognition unit (includes

EOG pre-processing part and EOG processing part),

and recognition results output part. EOG acquisition

circuit is used to acquire EOG signals and transmit

them to the computer via USB. EOG recognition unit

undertakes online noise removal and eye movements

detection. According to the pre-designed command

tables, the recognition results will be transformed into

the output control commands, at the same time, the

original EOG data will be displayed on the screen and

saved real-time.

2πRC (2)

where



is circumference ratio, we set the values of

the adjust resistance () and the capacitor () to

make ; the main amplifier circuit (B),

which uses some low-noise, high-precision and high

input impedance amplifier chips OPA2227, is designed

to complement the entire magnification required and the

formula for gain is

f =0.159Hz

G=1+

R (3)

3. SYSTEM IMPLEMENTATION

3.1. EOG Acquisition Module Then a low-pass filter (C) is used to remove power-

frequency interference and high frequency components of

EOG signals and the corresponding transfer function is

The design circuit of EOG acquisition circuit is shown in

Figure 4. From it we can see that the EOG acquiring

circuit is mainly composed of the emitter follower, the

pre-amplifier (A), the high-pass filter (B), the main am-

plifier (C) and low-pass filter(D). 22

434

H(s)= 3C Rs+CC R s+1 (4)

In order to suppress some interfere and isolate this

circuit form the other circuit, the emitter follower is

adopted, and then the pre-amplifier (gain=10) ampli-

fies the EOG signal to an appropriate amplitude. A

band-pass analog filter (0.159-10Hz) is used to remove

the base- line and higher frequency interference. After

main amplifier with 800 gains, the amplified EOG

signal is converted to digital signals and transmitted to

the computer.

According to Eq.4, the cut-off frequency will be set

by adjusting the resistance () and capaci-

tances ( and C).

f=10Hz

On the other hand, during the data acquisition process,

the computer will output some control signals to adjust

the state of the circuit, such as channel chosen, the sam-

pling rate, etc. The corresponding signal follow chart is

shown in Figure 5.

According to the characteristics of EOG signals, the

differential amplifier chip INA128 which is blessed with

low-power and high CMRR is used in pre-amplifier, its

gain can be adjusted by changing the value of the resis-

tance () and computed by the Eq.1.

3.2. EOG Signal Preprocessing

3.2.1. 50Hz Narrow Notch Filter

50Hz power-frequency interference makes some diffi-

culties and errors while the system detecting EOG sig-

nals. So, a notch filter is employed to remove 50Hz

power interference [3]. The notch filter can be designed

as follows:

50kΩ

G=1+R (1)

For removing DC drift and noise of EOG signals, a

high-pass filter (A) is employed and its cut-off frequency

can be computed by Eq.2.

)12(cos21

cos21

)( 









zbzb

bzH



(5)

HEALTH

OPA4227

R6 R5

OPA4227

INA128

Emitter Follower

EOG IN

OUT

Figure 4. EOG acquisition circuit.

42 Z. Lv et al. / HEALTH 1 (2009) 39-46

Figure 5. EOG acquisition fundamental block diagram.

The filter parameter b is expressible in terms of the 3-dB

width



 (in units of radians per sample) as follows:

)2/tan(1





b (6)

The Q-factor of a notch filter is another way of ex-

pressing the narrowness of the filter. It is related to the

3-dB width and notch frequency by:

Q00







 (7)

Thus, the higher Q, the narrower the notch. The trans-

fer function is normalized to unity gain at DC.

In this system, the EOG is sampled at a rate of 1KHz,

and the digital notch frequency will be:

sampleradians

/1.0

1000

502





 (8)

Designing a Q-factor of 50 for the notch filter, we

have a 3-dB width:

sampleradians

/002.0

/22 1

















(9)

Use the design Eq.6 to attain b=0.9969. Hence the

notch filter is:

2-1-

2-1

0.9937z1.8962z1

1.9021z1

9969.0)( 







zH (10)

The waveforms of original EOG signals and proc-

essed signals are shown in Figure 6.

Figure 6. (a) EOG signal with 50Hz noise, (b) EOG signal with 50Hz noise removed.

Filters

Amplfier

A/D

Interface

Control Circuit

USB

Choose

channnel

FIFO

Pre-am

lfie

EOG

Z. Lv et al./ HEALTH 1 (2009) 39-46 43

3.2.2. Using Dynamic Threshold to Normalize

The amplitude of EOG signals that produced by different

users is different, even if every blinking of a person is

different too. To avoid the problem of signal variability,

in our HCI system, a threshold method is employed to

transfer the EOG pulses into the square pulses for further

processing (we call it EOG pulse normalization). But the

small fluctuations on EOG waveform may bring some

troubles in EOG pulse normalization. We deal with the

problem using the dynamic threshold instead of tradi-

tional fixed threshold. Figure 7 interprets the funda-

mental of dynamic threshold.

In Figure 7, supposing the initial threshold is A, the

dynamic range is B. It’s obvious that three rectangle

pulses will generate if only the fixed threshold A is used

in the detection. In an approach using dynamic threshold,

once the system detects the first sample, the threshold

changes to A-B, then keeps this adjusted threshold until

the second sample is detected, and the threshold changes

to the initial value A again, vice versa. If the amplitude

of EOG signals is higher (lower) than the initial thresh-

old A, it is set to 1, otherwise, it equals 0. In this way,

some rectangular pulses can be acquired and waveforms

are shown in Figure 8.

3.3. Blinking Detection

3.3.1. Derivation

To count blinking in a specified time conveniently, nor-

malized signals should be processed derivation firstly.

The waveforms are shown in Figure 9. After derivation,

Figure 7. Fundamental of dynamic threshold.

blinking can be recognized easily and counted in a nu-

merical way.

3.3.2. Counting Blinking

When the system works, the program only detects the

positive pulses and ignores the negative ones. The posi-

tion of the first positive pulse is named “start_point”.

When the second pulse comes, its position is named

“end_point”. If the value(end_point-start_point) which

named DIF, is smaller than 1500(sample at a rate of

1KHz, so the time is 1.5s), the system will identify the

process as a blinking action, and the number of blinking

adds 1. When the third pulse comes, its position is

marked “end_point” instead of the foregoing value, and

then adjusts the DIF again. If it is smaller than 1500, the

program will continue to increase the number of blinking,

else it is recognized as two blinking actions and the

number of pulse is cleared. The flow chart of the soft-

ware is shown in Figure 10.

Figure 8. EOG signals after normalization.

Figure 9. EOG signals after derivation.

Initial threshold: A

First sample

Dynamic

range: B

EOG+50Hz noise

Adjusted

threshold: A-B

Second sample

44 Z. Lv et al. / HEALTH 1 (2009) 39-46

Figure 10. Software flow chart of blinking detecting.

3.4. Eye Movements Detection

An online detection algorithm is emphasized in our HCI

system. For the sake of an easy explanation, this paper

only analyzes the process of detecting vertical move-

ments, while the method of detecting horizontal move-

ments is same to it.

Start

3.4.1. Normalization

After the narrow notch filter removed the 50Hz power-

frequency interference, the original EOG signals (Fig-

ure 11(a)), managed by dynamic threshold (includes a

positive and a negative threshold), would be trans-

formed to a serial of rectangular pulses which have-1 or

1 in their amplitude, the waveforms are shown in Fig-

ure 11(b).

3.4.2. Getting Eye Movement Direction

Many experiments show that the phrase difference be-

tween the upward-rolling signals and the downward-

rolling is 180º. Hence, a series of rectangle pulses which

polarity is well-regulated are designed and named as

“Referenced Pulses” whose waveform is shown in Fig-

ure 12(a).

Multiply “Referenced Pulses” by “Normalized EOG

signals”, the system gets another series of rectangular

pulses, shown in Figure 12(b).

Compared with the waveforms in Figure 10(a), it’s

obvious that positive pulses mean rolling upward and

(a) Original EOG

(b) Normalization

Figure 11. EOG signals after normalization.

Initialization

DIF<1500?

End

First pulse position

Next pulse position

Blinking number++

Exit program?

Z. Lv et al./ HEALTH 1 (2009) 39-46 45

(a) Referenced Pulses

(b) Results

Figure 12. Get eye movements direction.

negative pulses mean rolling downward.

3.4.3. Judgment

The following method is adopted to judge the action of

eye movements. Firstly, an array(EOG_UD[1]) is de-

fined to record the polarity of two adjacent pulses. If

the value of the present sample is 0 and the foregoing

sample is 1, then EOG_UD[i]=1(i=0 or 1) and the sys-

tem will judge it as a positive pulse. On the contrary,

EOG_UD[i]=-1 and it is a negative pulse. Next, the

values of EOG_UD[i] should be checked for judging

the rolling direction. If EOG_UD[0] and EOG_UD[1]

are both 1, that means the user is rolling upward; and

so the situation of rolling downward.

4. EXPERIMENTS

The system has two input channels and the sampling rate

is 1000Hz, sampling precision is 16bit. In course of ex-

periment, the screen will play the original EOG signals

(like Figure 10(a)) and the result waveforms (like Fig-

ure 12(b)) real time, and these results are used to control

a remote mini-car to move via computer’s parallel port.

When the user rolls his eyes upward twice continually,

if he/she observes that the detection is correct in the

screen, he /she just blink three times quickly to confirm

the action, the program will attain a command to drive

the minicar. By contraries, if the detection is error, the

user closes eyes for about 3s to restart the system. The

direction of eye movements and corresponding actions

of the minicar are shown in Table 1.

5. CONCLUTIONS

Many applications can be developed using EOG because

this technique provides the users with a degree of inde-

pendence in the environment. Therefore, any improve-

ment in the convenience of this technique would be of

great potential utility in the future. If the eye movements

are known, various user interfaces can then be developed

to control different tasks: spell and speak software pro-

grams allow users to write a letter or a message, after

which a control system can interpret the message and

generate different commands to execute. A similar code

can be generated for deaf people, etc.

Table 1. Relationship between eye movements and output

commands.

Eye Movements Commands

Upward-Upward Go forward

Downward-Downward Go backward

Leftward-Leftward Turn left

Rightward- Rightward Turn right

46 Z. Lv et al. / HEALTH 1 (2009) 39-46

Experiences show that the system has a high-level sta-

bilization after long term testing. Non-linear error of A/D

converter is about () and power

wastage of supply is less than . The

program which uses VC++ 6.0 to achieve runs in win-

dows XP, it have not been found some fatal faults,

waveforms can show EOG signals exactly and real time.

LBS5.1mVLBS 44.21 

(5(500  VmA %))10

6. ACKNOWLEDGEMENT

The research work described in this paper is supported by nature sci-

ence foundation of Anhui province (60771033) and national nature

science foundation (070412038).

REFERENCES

[1] R. Barea, L. Boquete, M. Mazo, and E. Lopez, (2002)

System for assisted mobility using eye movements based

on electrooculography, IEEE Transactions on Neual

Systems and Rehabilitation Engineering, 10(4), 209-218.

[2] T. O. Ya, M. K. Asumi, (2005) Development of an input

operation for the amyotrophic lateral sclerosis communi-

cation tool utilizing EOG, Medical and Biological Engi-

neering, 43(1), 172-178. (In Japanese)

[3] S. J. Orfanidis, (1999) Introduction to signal processing,

Rutgers University, 407, 365-367.