Development of an Eye Tracker

1
Development of an Eye Tracker

• By Jason Kajon Barrett
• of the Center for Imaging Science
• at the Rochester Institute of Technology
• Advisor Jeff Pelz

2
What is an Eye Tracker?

• A device used to detect eye motion.
• What are they for ?
• Access for the physically handicapped,
• target variable resolution,
• realistic virtual reality environments,
• and increased bandwidth of human-computer
  interaction. Shannons Rule C BW log2(1
  S/N)

3

• Types of Trackers
• mechanical- corneal cover
• magnetic- eyeball polarity
• electric- EMG EOG
• optical- IR illumination

4
What is the Research Design?

• A device using four IR photodetectors and a
  microcontroller to detect the onset of rapid eye
  motion. A PicBasic program to run this device.

5
Why a microcontroller ?

• Simple data manipulation,
• lookup table,
• single upgradable chip,
• cost for device,
• and ease burden on CPU.

6
Targets for the Research

• lt 300 Project Expense

• 500 Hz Sampling Rate

• lt 4 Degrees Accuracy

7
(No Transcript)
8
Initial Budget

• Microcontroller and instructions 25
• Crystal 2
• Power Supply 110
• Comparator 7
• Diff. OP-AMP 5
• Potentiometers 20
• IR Emitters 67
• IR Detectors 45
• Total 281

9
Targets for the Research

• lt 300 Project Expense

• 500 Hz Sampling Rate

• lt 4 Degrees Accuracy

10
Why 500 Hz ?

• Saccades- rapid eye movements
• last 30 - 120 ms,
• cannot be within 100 ms of eachother
• video eye tracking unit at 60 Hz (i.e.
  17 ms, half a short saccade.)

11
Sampling Rate

• 200 hsec / instruction
• 200 instructions
• 40 msec for sample
• 50 kHz sample rate

For a 20 MHz PIC16C74A/JW-ND processor
12
Sampling Rate

• 20 msec / instruction
• 100 instructions
• 2 msec for sample
• 500 Hz sample rate

For a 10 MHz PIC16C84 processor
13
Targets for the Research

• lt 300 Project Expense

• 500 Hz Sampling Rate

• lt 4 Degrees Accuracy

14
Why 4 Degrees?
The fovea covers approximately a 10 Field of
View. Video eye trackers can come to within 1/2 0
for field of view. The main consideration is
speed.
15
To get within 4 Degrees

• individual calibration will be necessary
• extensive lookup table creation
• consecutive ON/OFF state comparison ( provided
  for by software )

16
Sample Calibration Target (with Nine
Targets)
17
Sample Calibration Target (with Seventeen
Targets)
18
The Proposed Design

• ASL Model 210 Eye Movement Monitor
• made from scratch
• single eye measurement
• fit within HMD

19
ASL model 210 eye movement monitor
20
(No Transcript)
21
Detector Positions as the Viewer Shifts Their
Horizontal Gaze
22
(No Transcript)
23
Detector Positions as the Viewer Shifts Their
Vertical Gaze
24
ASL video eye tracker
25
Long-term Goals of the Research

• Measurement of saccadic eye movements.

• Use in future projects involving eye contingent
  updating and the study of perception.

26
Conclusions
The eye tracker proposed is possible for the
proposed expense as there is an adequate
sacrifice for accuracy of eye position to
increase the speed. The resultant code written
in PicBasic will satisfy the design proposed.