Comments:
Summary:
StarGazer, a 3D pan/zoom interface for presenting noisy data is described. A two tier concentric arrangement of letters is used as a keyboard with which the system is tested. A rectangular area in the center of the display is used for zooming in on the display space. During use, a visual saccad results in a radial translation of the target area to the center of the screen where it is magnified. The greater space between objects at the center of the display increases accuracy and enables additional contextual information to be presented. Panning enables adjacent areas to be explored while maintaining the same level of magnification. Visual and auditory feedback are provided. Unlike other eye tracking preparations, StarGazer uses the gaze vector to determine where in 3D space the eye is looking.
Testing consisted of six preparations, three screen sizes with or without noise. Each of the 48 subjects were assigned one preparation. They were asked to type their name on a standard keyboard and the given an explanation of the StarGazer system. They were then given a two minute familiarization period with the system and finally they were asked to enter their first and last name as quickly and accurately as possible.
Two further tests were performed with 3 subjects to explore the effect of latency, and with 7 subjects to explore the effect of zoom.
Results are questionable as the testing procedures do not have consistent group sizes or adequate controls.
Discussion:
The StarGazer interface is a creative solution for overcoming the limitations of affordable eye tracking technology and providing a screen size independent solution.
However there are several points which are not satisfactory:
No description of the type of noise introduced is given.
No gaze direction vector is supplied by the eye tracking hardware, yet the use of this by StarGazer was identified as a novel attribute.
During testing, there is no control for individual differences as subjects were not required to perform all tests.
Furthermore, subsequent tests varying latency and zoom were conducted with disproportionately smaller subjects some having prior experience.
WPM comparison with another system showed StarGazer to be half as fast in its original form.
Robustness from noise attributed to low error rate could be due to other factors, i.e. concentration due to task difficulty or unfamiliarity etc.
A more comprehensive testing procedure with proper controls would give a more representative evaluation which could be compared to other similar systems.
I think this paper was a good introduction to this particular eye interaction method, and could be expanded upon in future papers for those points you mentioned. I also found many of the testing methods to be unsatisfactory, so further testing would be appreciated.
ReplyDeleteI agree that the paper skimped out on a couple of important details. The only details about the noise introduced was: a uniform noise within a 100 pixel radius with an update rate of 60Hz, which is vague.
ReplyDeleteI agree, more comprehensive testing was necessary. Also, attributing low error rate due to concentration on difficult task or unfamiliarity is quite possible, very good point.
ReplyDelete