1. Visualization of Eye Gaze Data using Heat Maps
Shijia Liu

2. Overview Background Existing Techniques Modifying Heat Maps
Using pupil’s horizontal and vertical coordinates
Fixation maps
Heat maps
Modifying Heat Maps
Visualization Options
Conclusion

3. Background Eye movements ↔ Internal brain processes
Window into human mind

4. Why the eyes’ movements is important?
From the collected eye-movement data, researchers can determine whether users were looking at the appropriate objects, whether the users are reading some particular words or phrases, learn the relative intensity of the user’s attention to various parts of a web page, and find out whether the user was searching for a specific item .

5. Techniques for Eye Gaze Visualization
Most straightforward technique:
Providing a simple plot of the pupil’s horizontal and vertical coordinates against time.
Then stimulus image as the background.

6. Techniques for Eye Gaze Visualization
More advanced visualization techniques -Fixation Maps
Convert the raw data into a set of fixations.

7. Techniques for Eye Gaze Visualization
Fixation Maps
Traditionally, fixations are
represented by circles
And saccades are
represented by lines connecting
the circles.

8. Techniques for Eye Gaze Visualization
A modern visualization technique – Heat Map
Was derived from the fixations maps.
Heat Maps
Fixation Maps

9. Techniques for Eye Gaze Visualization
Heat Maps
Heat maps better separate different levels of
observation intensity than fixation maps.
Color mapping is usually selected so that
the longer the observation, the warmer
color used to represent it.

10. Techniques for Eye Gaze Visualization
An existing modification of the technique uses opaque heat maps with no shadow over the unobserved areas . However, this kind of visualization hides details of the stimulus image. This tends to hinder the analysis.

11. Modifying Heat Maps
How to modify, the color ? Transparency?

12. Modifying Heat Maps
In this current implementation, the background can be partly or totally hidden by a shadow or fog.
Shadowing means covering the background picture in black, whereas fogging is covering it in white.

13. Modifying Heat Maps Transparency
Longer fixations add more transparency than shorter ones. A fixation longer than some threshold makes the display totally transparent at the location.

14. Modifying Heat Maps Fixations & Pixels
In fact, the measured gaze position(1 fixation) contains a group of pixels.
Radius

15. Modifying Heat Maps
Every pixel in the heat map related to a particular gaze location( fixation) extends the transparency to the neighboring area.
How could we distribute the transparency in one fixation?

16. Modifying Heat Maps Functions of Transparency Distribution: Linear
Linear+Sine Wave
Gaussian

17. Modifying Heat Maps Functions of Transparency Distribution:
In this figure, the x axis denotes distance from a fixation, and the y axis denotes influence on the transparency.

18. Visualization Options
There is a dialogue window contains the controls for setting the type of transparency and its distribution.

19. Visualization Option Sensitivity (S) denotes the radius of influence.
Brightness level (BL) is used as the gain for the distribution functions. It shows the minimum fixation duration (FDi) that makes the view totally transparent at the place of its occurrence.
Hiding level (HL) denotes the initial level of the shadow’s or fog’s opaqueness. Totally hidden background of the view corresponds to HL of 100%.

20. Visualization Options
Slicing
It means how do we want to slice or change the transparency
None
Steps
Mask
Relief

21. Visualization Options
Slicing Options
1.By default –None
Which means that transparency just changes by one of the distribution rules and no slicing is applied.

22. Visualization Options
Slicing Options
2.Steps
Transparency is shaped like terrain steps. The width of each step equals the value of Step level in pixels.

23. Visualization Options
Slicing Options
3.Mask
Similar to the 1st one(none).
With the option Mask checked, transparency lower than the value of Mask level is not applied to the visualization.

24. Visualization Options
Slicing Options
4.Relief
Similar to the 2nd one(Steps)
With the option Relief checked. This makes visualization similar to the topographic map.

25. Visualization Options
How do we get the value of Transparency for a pixel in one fixation?
Transparency Ti of a single pixel at distance D from the ith fixation in the Linear distribution is calculated using the following expression:

26. Visualization Options
For the Gaussian distribution, we keeps T = 100% when FD = BL for any value of S and BL:
Any Ti can not less than 0 or greater than 1.

27. Visualization Option Color Scheme
We can also add color scheme to modify the area with semi- transparent colors.
Blue – Green – Red – White;
Blue – Green – Red;
Blue –Red;
Green – Yellow – Red – White;
Green – Yellow – Red.

28. Conclusion
This technique is an extension of heat-map based visualization method. In this case, eye gaze is allowed to add transparency to the shaded background. This way the details of less relevance are hidden whereas more intensively observed areas become more conspicuous through increased transparency.

29. THANK YOU!