|M.Sc Student||Tomer Elbaum|
|Subject||Cyclopean eye vs. Dominant eye in using eye input interface|
|Department||Department of Industrial Engineering and Management||Supervisors||Full Professor Erev Ido|
|Dr. Michael Wagner|
|Full Thesis text|
Gaze interface is the use of human gaze to control computer systems. In this study we offer a human factors perspective for coping and reducing gaze interface inaccuracy problem. The inaccuracy problem is the experienced difficulty of its users to accurately interact with computerized objects. The study explores the possibility of using the cyclopean eye, the midpoint of left and right eye positions, or alternatively using the dominant eye, for controlling gaze interface systems. The study focuses on the improving effect of these two possibilities on gaze interface accuracy. By integrating a well-established phenomenon termed fixation disparity with the cyclopean eye theory, we suggest that left eye and right eye often inaccurately fixate on an observed target and "land" at both of its sides. Instead, only the cyclopean eye frequently “land” closer to target. Although this suggestion is challenged by the dominant eye theory which stems from brain laterality characteristics, we hypothesized that gaze interface’s cursor should be controlled by the cyclopean eye rather than by the dominant eye.
In Experiment 1, 27 participants performed free (no gaze interface) target tracking task. We hypothesized that cyclopean tracking will be more accurate than dominant and non-dominant eye (Hypothesis 1). Findings supported our hypothesis, showing that target-to-eye deviation was smallest for the cyclopean eye, demonstrating best tracking performance. In Experiment 2, the same 27 participants performed a gaze interface target tracking task. We hypothesized that compared to dominant and non-dominant eyes, gaze interface tracking performance will be optimal when cursor is controlled by the cyclopean eye (Hypothesis 2). We also hypothesized that cyclopean eye will demonstrate increase tracking superiority in higher levels of fixation disparity (Hypothesis 2a). Findings supported both our hypotheses. When cyclopean eye controlled the cursor, the percent of time on target was higher, compared to dominant and non-dominant eye cursor control. Also, according to hypothesis 2, the percent of time on target was highest for cyclopean-cursor-control during high level of fixation disparity.
is controlled by the cyclopean dominant eye tracking accuracy was close to cyclopean in the second half of the experiment.
From a theoretical viewpoint, our findings primarily correspond with the cyclopean perspective and may partly extend its predictions about egocentric direction to more dynamic environments. It also indicates possible motor advantage of the dominant eye. From a practical perspective, our findings suggest that using a single eye to control gaze interface often leads to greater inaccuracy and misrepresents users’ intention. A key strength of the present study is its implications for gaze-interface ventures, manufacturers and software developers. First, we recommend using a binocular, rather than monocular eye trackers, as the infrastructure for gaze interface systems. Second, we strongly recommend using the cyclopean eye, rather than a single dominant eye, as the controller for gaze interface systems in order to improve accuracy and system performance and to better accommodate users’ intentions. This recommendation becomes essential in cases of high fixation disparity levels.