Electrooculography Human Computer Interface

EOG HCI
Electrooculography Human Computer Interface
Introduction
• Currently, 12 million people within the U.S require assistance for
ordinary tasks. While computer technology can help individuals with
upper body movement disabilities independently handle a wide range
of activities, they cannot provide computer inputs with a standard
keyboard or mouse. The purpose of this project was to provide these
people with a cost-effective control system for home computers.
• Electrooculography (EOG) is a technique that measures steady state
corneal-retinal potential of the eye balls. EOG signals are collected by
two pairs of surface electrodes placed around the eyes. Compared
with other biological signals, they have larger amplitude and
distinguishable patterns for different types of eye movements.
Furthermore, the signals linearly change as eyes rotate away from the
center.
System Architecture
Top Level Architecture
PC Monitor
• User wears EOG goggles.
• A single cable from goggles
connects to The EOG HCI.
• USB cable from EOG HCI
connects to PC.
• Mouse cursor movement is
visible on PC Monitor screen.
• Cross-platform compatibility.
PC Video Output
PC
Electrode Cable
USB Cable
User
wearing EOG Apparatus
EOG HCI
Level-2 Architecture
5V Power
Power Filtering
User wearing
EOG HCI Spectacles
Multi-Channel ADC
Power Convertion
Raw EOG Signal
Digital Horizontal
and Vertical Values
Signal Amplification
• Taking advantages of EOG signals, this project developed a low-cost
and power-efficient device, “Eye-Mouse” ,which used eye movements
to control joystick-like movements of the computer cursor in the four
cardinal directions.
Mouse Movement
Algorithm
Analog Difference Signal
(between 0V to 1V)
Signal Filtering
Mouse Movement
Speed and Direction
Signal Differentiation
HID Compliant Mouse
Movement
Signal Offset
USB
Implementation
Results
Using 3D printer, a set of adjustable goggles
were designed and fabricated to house the
EOG electrodes. Afterwards, a circuit was
designed to provide a differential EOG
signals on the needed range by the ADC
(Analog to Digital Converter) on the
microcontroller unit. The circuit was
prototyped on breadboard, Vector board and
PCB.
Horizontal Movement
COMPUTER
SAM4S ARM MCU
CIRCUIT
A testing application was designed based on Fitts’s Law.
Performance was compared to two other conventional mice:
Logitech Laser vs Microsoft Optical vs EyeMouse EOG
Comparison to Minimum Fitts's Law
Requirement
5000
4500
Logitech Laser
Microsoft Opticle
4000
Time (ms)
3500
Microsoft Opticle
Logitech Laser
R² = 0.264
3000
Eye-Mouse EOG
2500
2000
1000
Following signals were fed to the ADC on the
vertical and horizontal channels.
0.00
0.10
0.20
0.30
0.40
0.50
0.60
0.70
0.80
R² = 0.1522
Linear (Logitech
Laser)
500
0.90
0
R2 Value
Performance of Mouse
Linear (Microsoft
Opticle)
R² = 0.7743
1500
Eye-Mouse EOG
0
1
2
3
Linear (Eye-Mouse
EOG)
Index of Difficulty
Minimum Requirement of Fitts's Law
Vertical Movement
Movement time was compared against two other conventional mice
and averages of consecutive movement between the same distance
were recorded:
Comparison of Average Time for 13 Trials
Logitech Optical vs Verbatim Laser vs EyeMouse EOG
Time (ms)
35000
Logitech Optical
30000
Logitech Optical
25000
Verbitim Laser
Verbatim Laser
20000
Since Eye-mouse is detected as a generic USB mouse, it is multiplatform compatible.
2066.769231
Eye-Mouse EOG
y = 680.68x + 7331.4
15000
Linear (Logitech
Optical)
10000
Digitized signals then are processed on the microcontroller to
determine the eye movements. An algorithm was setup for blinks,
horizontal, and vertical eye movements to associate them with
joystick-like movement of the mouser cursor on the screen.
3365.307692
5000
y = 10.049x + 3295
0
y = -2.2857x + 2082.8
5
10
0
Eye-Mouse EOG
Linear (Verbitim
Laser)
15
Linear (Eye-Mouse
EOG)
Trial
12096.15385
0
2000
4000
6000
8000
10000
12000
Time (ms)
Average Time for Movement
Movement
Result
Comparison to Conventional Mice
Horizontal
0.264 linear correlation co-efficient
41.1% less performance
Vertical
12096 ms average movement time
77.5% slower movement
14000