Open Access
Refine
Year of publication
- 2015 (1)
Document Type
- Master's Thesis (1) (remove)
Language
- English (1)
Has Fulltext
- yes (1) (remove)
Is part of the Bibliography
- no (1) (remove)
Keywords
- Eye tracking (1) (remove)
Institute
- Optik und Mechatronik (1) (remove)
Purpose
Automated scanpath comparison metrics should deliver an objective method to
evaluate the similarity of scanpaths. The aim of this thesis is an evaluation of
seven existing scanpath comparison metrics in static and dynamic tasks in order
to provide a guidline that helps to decide which algorithm has to be chosen for a
special kind of task.
Methods
The applicability of the algorithms for a static, visual search task and a dynamic,
interactive video game task as well as their constraints and limitations were tested.
Therefore, binocular gaze data were recorded by using the eye tracking system The
Eye Tribe (The Eye Tribe ApS, Copenhagen/ Denmark). Objective task performance
measures from 21 subjects were used in order to create scanpath groupings
for which a relevant effect of dissimilarity was to be expected. Objective task performance
measures such as task performance time were statistically evaluated and
compared to the results gained by the comparison metrics.
Results
Four of the algorithms being used successfully identified differences for static and
dynamic tasks: MultiMatch, iComp, SubsMatch and the Hidden Markov Model.
ScanMatch was very sensitive for the static task but not applicable to the dynamic
task whereas FuncSim was suitable for dynamic but not for static tasks. Eyenalysis
failed to detect any effect.
Conclusion
The applicability of scanpath comparison metrics depends on the state of the task,
respectively on the kind of experimental set up. In future, the application area for
eye tracking will expand and an improvement of automated scanpath comparison
metrics is therefore required.