RMC Symbiotic Computer Lab Presentation

1
RMCSymbiotic Computer LabPresentation
An Experiment Using Eye Trackers
Maj JW Paul, RMC
2
Caveat

• The following project was carried out as part of
  CISC839
• 6 week course 8 week project
• My goal
• play with eye tracker technology
• wanted an experiment vice build something
• see tarpit.rmc.ca/paul/

3
An Investigation of Foveal and Peripheral Vision
During Repetitive Tasks
Maj JW Paul, RMC
4
Outline

• Background Idea
• Research
• The Experiment
• The Results
• Lessons Learned

5
Background
6
The birth of an idea..

• Cockpit overloading
• Can eye-tracking help?
• problem looking without seeing
• Look at small subset of flight regime
• Instrument Flight Rules (IFR) flight

7
CF5 - Instrument Panel
8
CF5 - Instrument Panel
IFR scan will depend on flight regime
9
Types of Errors

• Slips (right intention, incorrectly executed)
• subconscious actions intended to satisfy our
  goals get waylaid by automatic behavior
• rule-based or skill-based
• Mistake (wrong intention)
• Conscious deliberation leads to incorrect
  generalizations or classifications
• rule-based or knowledge-based

also Lapses - right intention, not executed
10
What we know
11
Terminology

• Foveal vs Peripheral
• Para-Foveal vs Extra-Foveal
• Focal vs Ambient
• Saccade

12
Peripheral Ability
Search task with moving stimula -slow data input
rate -circles on paper strip -searching for
squares
From Mackworth (1996)
13
Difficulty of Foveal Task (on peripheral vision)
100
No Task
90
Adding two single digit numbers
80
Adding three single digit numbers
70
Adding four single digit numbers
60
Percentage correct
From Murata and Miyoshi (2000)
50
40
30
20
10
0
5
15
25
35
Offset in degrees
14
Difficulty of Peripheral Task
100
Detecting line segment
90
Orientation of U
80
Idenitification of letter (H L O X)
70
Classifying three letter word
60
Percentage correct
From Edwards and Goolkasian (1974)
50
40
Note - Noise Effects as well Mackworth
(1965) visual noise causes tunnel vision Mori and
Hayashi (1995) Visual Interference with Users
Task on Multiwindow Systems
30
20
10
0
0
20
40
60
Offset in degrees
15
What we want to know
16
Hypothesis

• As repetitive nature of task increases, so will
  time taken to notice a peripheral event
• As repetitive nature of task increases, more
  peripheral events will be missed
• As repetitive nature of task increases, so will
  time taken to notice a foveal event
• As repetitive nature of task increases, more
  foveal stimula will be missed

17
What are we measuring

• Dependant variable
• time of task
• location of event (peripheral vs foveal)
• rate of event input (vigilance task)
• Independent variable
• miss rate
• identification time

18
Planning TheExperiment
19
Experimental Display
17 monitor
Theoretical Field of View User 20 from screen
20
Experimental Display
CF5 Instrument Panel
21
Experimental Display
next - remove noise
Scan Pattern
22
Experimental Display
next - normalize
23
Proposed Display
Scan Pattern
24
Initial Display
25
Running TheExperiment

• Nothing is ever a total failure,
• It can always serve as a bad example.

26
Display Mark II
27
Results - mean by subject
600.0
500.0
400.0
Msec Delay
300.0
200.0
100.0
0.0
8
1
2
3
4
5
6
7
Base-F
Subject
Base-P
Scan-F
Scan-P
28
Statistical Analysis

• Foveal lt Peripheral (p.004)
• Baseline lt Scan (p.001)
• H1 foveal(t) increases with t (p.0006) ?
• H3 peripheral(t) increases with t (p.0020) ?
• H2 H4 missed events increase with t ? note -
  some events initiated on saccade

29
Lessons Learned

• Literature Survey
• Advice

• Measure EVERYTHING
• Do one to throw away

30
Questions?