Date of Award
Campus Access Dissertation
Doctor of Philosophy (PhD)
Eye tracking research is essential in certain domains, such as visual perception, human-computer interaction, or computer vision. Despite the convenience of affordable, efficient, and “claimed to be accurate” eye trackers, there are several things to be considered when it comes to developing the eye tracking experiment and analyzing the eye movement data. Here, a study on the characteristics of raw gaze data, how to extract features from the gaze data with necessary considerations, and interdisciplinary cognitive science research is presented.
The current eye tracking systems’ accuracy is not always ideal. Moreover, even with a reasonable calibration result, the accuracy of gaze tracking deteriorates over time. It was found that the offset between the reported gaze positions from the eye tracker and the desired fixation target on the display was nearly constant across the screen. Thus, an algorithm that maintains gaze tracking accuracy by compensating for those offset vectors was developed.
For the feature extraction from eye movement data, important considerations in a visual attention study using simple pro- and anti-saccade tasks on three different groups (Alzheimer’s disease, Mild Cognitive Impairment, and Control) of participants are discussed. The preliminary observations on eye movements features indicated that there was a distinction between groups in terms of eye movement features.
Finally, an interdisciplinary cognitive-behavioral research using the eye tracking system, EEG, and fNIR is presented. The purpose of the research was to explore the effectiveness of two-dimensional (2D) vs. three-dimensional (3D) representation in Science, Technology, Engineering, and Math (STEM) education. Specifically, the main focus was to study how swiftly and accurately a diverse group of community college students can compare models of molecules (a simplified chirality task) when those models were represented as 2D vs. 3D. Moreover, the behavioral differences between low- and high-performing observers was investigated. Some important findings from our preliminary analysis were 1) eye movement features were correlated with neuro-behavioral measures, 2) eye movement features reveal separation between poor and good performers, and 3) especially the binocular disparity feature reflects subjects’ performance on both 2D and 3D tasks.
Koh, Do Hyong, "The Study of Visual Attention: From Raw Data to an Interdisciplinary Environment" (2019). Graduate Doctoral Dissertations. 473.