Popelka, S: Space-Time-Cube for Visualization of Eye-tracking data
1. Space-Time-Cube
for Visualization of Eye-tracking data
Stanislav POPELKA
This presentation is co-financed by the
European Social Fund and the state
budget of the Czech Republic
2. Introduction
Eye-tracking is one of the methods of usability studies and is
considered as an objective
The modern eye-trackers use contactless measurements of the
visible parts of the eye and corneal reflection of direct beam
of infrared light
The reflected light is recorded by camera
From analysis of the changes of corneal reflection, the point of
regard is calculated
First InDOG Doctoral Conference, 29th October - 1st November 2012, Olomouc
4. The human eye performs several types of movement - the
most important are fixations and saccades
Qualitative information about eye movements describes the
way in which the user explores the stimulus
It can reveal areas of greatest interest, disruptive elements or search
tactics during answering the question
Quantitative information can be derived from eye-tracking
data through metrics of fixation and saccades
For example - the fixation length, saccade amplitude, fixation/saccade
ratio or AOI (Area of Interest) dwell time
First InDOG Doctoral Conference, 29th October - 1st November 2012, Olomouc
5. Visualization of eye-tracking data
HeatMaps and ScanPaths are common visualization methods
of eye-tracking data
They cannot effectively express the change of time
The cause of this problem is displaying of the three-
dimensional data (X, Y, time) in two-dimensional space (X, Y)
It is necessary to use spatio-temporal visualization
First InDOG Doctoral Conference, 29th October - 1st November 2012, Olomouc
6. Spatio-temporal visualization
Kraak and Ormeling (1996) describes three approaches to space-
time visualization
Simple static map (ScanPath)
If the static map displays complex time phenomena, it is likely that
the phenomena will overlap in the map, which can lead to loss of
information.
Series of static maps (ScanPaths)
It allows viewing changes in a phenomenon in several time periods
For huge series of static maps, the interpretation should be difficult
Animation (of ScanPaths)
Animation captures the dynamics of the space-time phenomenon
appropriately
Displaying of the development of the phenomenon comprehensively
for the entire study period is not possible
First InDOG Doctoral Conference, 29th October - 1st November 2012, Olomouc
7. Space-Time-Cube
Space-Time-Cube displays spatial and temporal
component at the same time
Space–Time–Cube is the most important element in
the Hägerstrand’s spatio-temporal model
Space-Time-Cube displays the map at the base of the
cube (axes X and Y) while Z axis is used to represent
time
Spatial and temporal components are shown
together, and relationship between space and time
can be revealed
First InDOG Doctoral Conference, 29th October - 1st November 2012, Olomouc
8. Space-Time-Cube
First InDOG Doctoral Conference, 29th October - 1st November 2012, Olomouc
9. Software for STC visualization
Visual Analytics Toolbox (CommonGIS)
Developed by Fraunhofer Institute IAIS
Gennady and Natalia Andrienko
GeoTime 5
Commercial application for Space-Time-Cube
Designed for geographical data
Direct import from ArcGIS
Space-Time-Cube extension for open-source GIS uDig
ITC in Eschende, Netherlands
Team around professor Jan-Menno Kraak
Unavailable at the moment
Extended Time-Geographic Framework Tools
Extension for ArcGIS 9.3
compared to the above mentioned applications, this extension has less
functionality
First InDOG Doctoral Conference, 29th October - 1st November 2012, Olomouc
10. Case study
Research of map reading during solving the geographic problems
Study was focused to the use of map legend
Students project – students of Masters program
Total of 16 respondents
8 cartographers (KGI students after cartography course)
8 non-cartographers (zoologists, lawyers..)
Total of 19 stimuli
Maps from school atlases
Unlimited time to read the question
45 seconds to answer
Short questionaire after the test
First InDOG Doctoral Conference, 29th October - 1st November 2012, Olomouc
11. Case study
During the research, SMI RED 250 eye-tracker developed by
SensoMotoric Instrument was used
The device allows data acquisition with frequency of 120 Hz
Point of regard of the eye, expressed with Y and Y coordinates
are recorded and stored with a regular interval of 8
miliseconds
(X, Y, t)
X – location
Y – location
t – time
…
First InDOG Doctoral Conference, 29th October - 1st November 2012, Olomouc
12. Results of case study
Respondents automatically looks into the legend first
Regardless belonging to the Cartographer/NonCartographer
group
If the legend is barely legible, they spend much more
time in it
One of the stimuli was a gimmick – respondents were
asked to find the coal mine in the map, but the symbol
was missing in the legend
First InDOG Doctoral Conference, 29th October - 1st November 2012, Olomouc
15. Space-Time-Cube example from the case study
Space-Time-Cube was used for visual analysis of users
interaction with maps from school atlases
Respondents were asked to quickly find areas where
the flax is grown and identify it by clicking the mouse
in the map
First InDOG Doctoral Conference, 29th October - 1st November 2012, Olomouc
17. Space-Time-Cube in Common GIS
Data modification is necessary before importing into
CommonGIS
Two types of visualization of measured data in a Space-
Time-Cube have been tested
Visualization of trajectory made directly from raw data
Visualization of fixations connected with lines (representing
saccades)
CommonGIS has not an ability to differ fixations based on
their size
First InDOG Doctoral Conference, 29th October - 1st November 2012, Olomouc
19. Space-Time-Cube in GeoTime
Occulus GeoTime 5 is the software for visual analysis of
time series data
Possibility of connecting GeoTime with ArcMap or
Microsoft Excel
Fixations can differ based on their lenght
Geotime has a possibility to analyze data – find
patterns, clusters, gaps, intersections in space and time…
Problems with coordinate system, units
Geotime is designed for geographical data (WGS 84)
Eye-tracker produce data in Cartesian coordinate system
(1680*1050 px)
GeoTime is unable to load data in miliseconds
First InDOG Doctoral Conference, 29th October - 1st November 2012, Olomouc
23. Future plans
Solve the problems with GeoTime
Use its functions for analysis
Use Space-Time-Cube for visual analysis of
results from next eye-tracking tests
Comparison of 2D vs. 3D visualization
First InDOG Doctoral Conference, 29th October - 1st November 2012, Olomouc
24. Thank you for your attention
Stanislav Popelka
standa.popelka@gmail.com
www.geoinformatics.upol.cz/ET
First InDOG Doctoral Conference, 29th October - 1st November 2012, Olomouc