An Electrooculogram Based Control System Md. Moin Uddin Atique*, Sakhawat Hossen Rakib, Khondkar Siddique-e-Rabbani Biomedical Physics & Technology University of Dhaka, Dhaka, Bangladesh E-mail:
[email protected]*,
[email protected],
[email protected], Abstract— Electrooculography (EOG) is the electrical activity due to eyeball movement, recorded from the different points on the face of a person. Human-Machine Interface (HMI) application based on EOG could be very helpful for the patients suffering from neural diseases like Amyotrophic Lateral Sclerosis (ALS) or paralysis. In the present work a low-cost device is developed to detect and switch on or off any of 16 LEDs in a 4×4 matrix. Two EOG amplifiers have been used to detect five different movements or actions of the eyes of a user. Right, Left, Up and Down movement of the eye shifts the position of the lit LEDs while a Blinking action of the eye activates an action on that chosen LED. The scheme has been implemented using an Arduino Uno microcontroller board. This mechanism can be used in a virtual keyboard or to control machines including a wheel chair by a paralysed person. In future this technique can be used in other HMI based applications as well. Keywords - Electrooculogram, EOG, Human Machine Interface, Automatic Machine control, Medical Instrumentation.
I.
INTRODUCTION
Electrooculography (EOG) is the electrical potential [1] that is produced across the cornea and retina due to eyeball movement of a person [2-3]. This potential for different movements of eye can be detected from the forehead, temple and upper side of the cheek. EOG can be used for HMI (Human-Machine Interface) applications. A controlling system based on this electric signal could be very helpful for patients suffering from neural diseases like Amyotrophic Lateral Sclerosis (ALS) [4], a type of Motor Neuron Disease (MND). ALS patients suffer from muscle weakness and shrinkage and lose the ability of movement and speaking. However, eye movement is not usually affected and EOG signals received from the ocular movement may provide ALS patients with a means to control equipment or communicating with others [5]. Many research teams around the world are working on building HMI system depending on EOG. Rafael Barea, along with his colleagues, developed a standard electric wheelchair with an on-board computer sensors and a graphic user interface that is eye-controlled [6]. Their system uses the eye movement as the mouse controller of a computer. This system also gives audio-visual feedback to the user to make the system user friendly. Some other research teams are developing EOG based graphical interface systems where patients will be able to communicate using only eyeball movement [7-8]. Some researchers used neural network to
recognize the EOG gesture. Controlling Robots and developing Human-robot interface are being attempted by many groups around the world [9-11]. A hospital alarm system was also developed, which is useful for emergency conditions [12]. Cost effective designs could be helpful to make the device available for people with low income [13-14]. The goal of the present work is to design a low-cost EOG based controlling system which will be useful for full body paralyzed patients, who can only move their eyes. EOG detection is difficult due to its very low amplitude and frequency. Developing a controlling system and providing multi functionality using only limited number of eye movements is hard to achieve. In the present work eye movement in four directions combined with blinking action have been used to choose and turn on or off any of 16 LEDs arranged in a 4×4 matrix, which can be the basis of applications like virtual keyboard for communication, wheel chair control for easy movement or any human machine interface applications. II.
METHODS & MATERIALS
Detection and utilization of EOG signal is a step by step procedure involving both hardware and software interfaces. These steps and procedures are described below.
Figure 1: Detection method of the EOG signal.
A. EOG detection and controlling system: For every movement of the eye an EOG signal is produced. EOG signal has very low potential difference of about 0.1 to 2 mV with a typical timespread of about 300ms to 400ms. The frequency range of this signal varies from Dc to 30 Hz. In this project an attempt will be taken to detect five basic movements of the eye: Left, Right, Up and Down movement of the eyeballs and Blinking action of the eyelid. This process requires two EOG amplifiers as different movement produces different electric potential across different points. The signals were collected using skin surface electrodes from the face. To detect the right and left movement of the eyeball, inputs of one of the amplifiers (amplifier A in Figure 1) are connected to electrodes placed on the temples of the two sides of the face (positions P and Q). This amplifier shows a positive peak of potential difference more than a threshold, +Va1, for eyeball movement towards right and negative (less than −Va1) for the left. The detection of Up, Down and Blink actions requires placing of input electrodes of the second amplifier (amplifier B) at the corner of the forehead and at the upper side of the cheek (position R and S in Figure-1). The common electrode for both the amplifiers is connected in the middle of the forehead (position T). This second amplifier will also produce positive potential above a threshold, +Vb1 and a negative potential less than −Vb1 for corresponding upward and downward movements of the eyeballs. In addition, a higher positive potential (more than +Vb2) will be found from this amplifier for blinking the eye which is distinguishable from the Up move. Double blinking signal could be distinguished by detecting a positive potential of more than +Vb2, twice within 250ms. Two or three such consecutive blinking signals may be detected as Double Blink or Tripple Blink to produce a confirmatory signal eliminating actions due to chance blinking. The left hand part of Figure 2 shows the scheme of the LED matrix and of the movements while the right hand part shows the Truth table for the whole detection procedure from both the amplifiers (n/a means not allowed).
In this case the user will be able to blink any selected LED from a two dimensional matrix (4×4) of 16 LED arrangement. The movements will be divided into two different types, 1. Shifting Commands (LEFT, RIGHT, UP DOWN) 2. Action Commands (BLINK, DOUBLE BLINK) To illustrate the mechanism, suppose initially LED number 6 is on (Figure 2). One of the four adjacent LEDs will be lit depending on the detected shifting command. For example, if Right movement is detected the LED 6 will be off and LED 7 will be on. In this way through the shifting commands the user will reach the target destination LED. Next a Blink action (or a Double Blink) will confirm the choice of the destination and for carrying out desired control action. B. Amplifier Design: EOG signals have potentials between 0.1 and 2 mV and a frequency range from DC to 30 Hz. The amplification required for typical applications should be about 1500 and a band pass filter in the range of 0.1Hz to 30Hz. To remove the common mode noise, which is mainly from mains power line of 50Hz, the CMRR of the amplifier should be about 80dB. The instrumentation amplifier in the present work was designed to have a gain of about 500. A high pass filter with a cut-off frequency of 0.1Hz, followed by a low pass filter with cut-off at 30Hz was used. A non-inverting amplifier was used with a gain of 3 which gave the desired total gain of about 1500. A level shifter was used to raise the signal level so that there are no negative potentials at the output. This is required because the signals will be detected using the microcontroller’s ADC (Analog to Digital converter) port which detects only positive signals.
Figure 3: Experimental Setup.
Figure 2: Movement scheme in the LED matrix and the Truth table for detection of various movements from EOG.
The amplified signal is to be analyzed properly by the microcontroller to distinguish different patterns of potential variation for different movement actions of the eye (e.g., left, right, blink) given in the truth table. This study of detection will be used to control any system with appropriate algorithm.
C. Hardware: To show that the controlling system works, a 4×4 matrix of 16 LEDs is used. The whole experimental setup is shown in Figure-3. An Arduino Uno board with a built-in microcontroller was used to develop the hardware. Eight pins from the Arduino (pin no. 4-11), connected to rows and columns of the LED matrix respectively, were used for the LED control. Two analog inputs (A0 and A1) were used to
acquire signals from amplifiers A and B. The power was provided using two 6V Lead-acid batteries. For the prototype, disposable self-adhesive ECG electrodes were used as shown in Figure 3. D. Software: The outputs from the two amplifier circuits are analysed to detect and distinguish different EOG signals. The scheme of detection is based on the observation that the EOG signals have a typical spread between 300ms and 450ms (shown in Figure 5). Instead of measuring the peak values we used a time difference method to detect the presence of the EOG signal. We keep on measuring the differences of consecutive output values at intervals of 50ms. When there is no EOG signal, this difference value will be low. As soon as an EOG appears, this difference will be high. From actual observation on a subject, thresholds for these difference values will be adjusted (to correspond to thresholds based on Va1, Vb1 and Vb2 respectively). A specific signal will be said to be detected if the corresponding potential difference value (at a time gap of 50ms) is greater or lower, as appropriate, than respective threshold values. If a signal is detected from amplifier A, Right movement will be indicated for signals greater than Va1 and Left movement will be indicated for signals less than –Va1. A steady value in the signal will be the indication for no movement. Start
Input
B
Signal Detected?
Signal B
Figure 5: Detected EOG signals (A: Right, B: Left, C: Up, D: Down, E: Blink, F: Double Blink)
III. RESULTS The outputs from the amplifiers are collected using a digital oscilloscope. The measurements are performed with the voltage axis set at 1 volt/div and time axis at 250 ms/div. A set of acquired signals are shown in Figure-5. Here, image A is indicated for Right movement of the eyeball, B for Left movement, C for Up movement, D for Down movement while E and F are indicated for the Blink and Double Blink actions of the eyelid.
n\a
Signal A A
B>+Vb2
+Vb1
