Rochester Institute of Technology
Gaze Guidance, Task-Based Eye Movement Prediction, and Real-World Task Inference using Eye Tracking
Abstract
dc:description.abstractThe ability to predict and guide viewer attention has important applications in computer graphics, image understanding, object detection, visual search and training. Human eye movements provide insight into the cognitive processes involved in task performance and there has been extensive research on what factors guide viewer attention in a scene. It has been shown, for example, that saliency in the image, scene context, and task at hand play significant roles in guiding attention. This dissertation presents and discusses research on visual attention with specific focus on the use of subtle visual cues to guide viewer gaze and the development of algorithms to predict the distribution of gaze about a scene. Specific contributions of this work include: a framework for gaze guidance to enable problem solving and spatial learning, a novel algorithm for task-based eye movement prediction, and a system for real-world task inference using eye tracking. A gaze guidance approach is presented that combines eye tracking with subtle image-space modulations to guide viewer gaze about a scene. Several experiments were conducted using this approach to examine its impact on short-term spatial information recall, task sequencing, training, and password recollection. A model of human visual attention prediction that uses saliency maps, scene feature maps and task-based eye movements to predict regions of interest was also developed. This model was used to automatically select target regions for active gaze guidance to improve search task performance. Finally, we develop a framework for inferring real-world tasks using image features and eye movement data. Overall, this dissertation naturally leads to an overarching framework, that combines all three contributions to provide a continuous feedback system to improve performance on repeated visual search tasks. This research has important applications in data visualization, problem solving, training, and online education.
Degree
thesis:*- Name thesis:degree_name
- Computing and Information Sciences (Ph.D.)
- Level thesis:degree_level
- Dissertation
Author and committee
dc:creator, dc:contributor.*- Author dc:creator
-
- Sridharan, Srinivas
- Contributors dc:contributor
-
- Reynold Bailey
Subjects
dc:subject × 6Identifiers
dc:identifier.*- Repository record dc:identifier
- https://repository.rit.edu/theses/9164