Research in eye tracking has been constrained by the high costs and reliance on specialised hardware. However, recent advances in technology, including progress in computational gaze estimation and the capabilities of front-facing cameras on handheld mobile devices, are enabling pervasive eye tracking, which is opening up new possibilities. This includes empowering users with gaze as an input modality for interacting with handheld mobile devices. While prior work investigated gaze interaction techniques in non-mobile, non-handheld settings, such as desktops, mobile devices held by a mount, or head-mounted displays, mobile contexts pose unique challenges due to mobile devices’ dynamic use, distinguishing them from the other previously explored settings. Therefore, it is not expected that results from prior work will be transferable to the mobile context, suggesting a lack of empirical work to evaluate and enhance the performance of gaze interaction techniques in this context. This thesis first evaluates the performance of gaze interaction techniques in a mobile setting. This also includes evaluating a combination of interaction techniques within a single interface to facilitate navigation and selection tasks. Second, the thesis focuses on exploring and addressing factors that degrade the performance of Dwell time in the mobile context, given its potential for target selection in the mobile context compared to other gaze techniques, such as Pursuits and gestures. This includes exploring target sizes while accounting for the dynamic nature of mobile use, which inherently involves varying distances between users and their phones. It also investigates an adaptive approach to mitigate the impact of the mobile context on Dwell time performance. Finally, this thesis also investigates the effect of encumbrance on gaze input, Dwell time, in particular. This marks the first step towards exploring the potential of using gaze as an alternative modality to touch when its performance degrades due to situational impairments. Experiment in Chapter 2 compared three of the most common gaze interaction techniques in the mobile contexts, Dwell time, Pursuits, and gestures, given a varying number of targets. Participants were freely interacting using gaze, while sitting and on the move. While the results showed that Pursuits was generally faster compared to other techniques, the performance and preference for the techniques varied depending on the context and the number of targets. Chapter 3 evaluated the concept of pairing input techniques in the same interface for target acquisition and navigation tasks, to facilitate the interaction, where one technique is used for selection, and another for navigation. When evaluated while participants were sitting and walking, results showed that pairing techniques can effectively enhance the interaction. Combining gaze techniques for interaction was generally well perceived by participants, compared to using one gaze input for both target acquisition and navigation tasks. Results from both experiments in Chapters 2 and 3 showed that participants indicated a higher preference for the Dwell time input technique. Following this, Chapters 4 and 5 explored and employed approaches to mitigate the impact of the target size on Dwell time when used in the mobile context. Experiment 4 specifically examined factors that degrade Dwell time performance by exploring the interplay between target size, viewing distance, and target region on tracking accuracy. The study identified targets of size 4° as an optimal fixed target size across viewing distances. However, handheld mobile use is inherently dynamic, as users change posture or move, even an optimal fixed size is perceived differently at different distances, which can degrade accuracy. Therefore, Chapter 5 explored an adaptive approach to mitigate the impact of dynamic mobile use on dwell performance. The novel approach updates target sizes in response to changes in viewing distances. A usability study compared the adaptive approach to using three static user interfaces (small, medium, and large). Results showed that the adaptive user interface leveraged the advantage of its distanceresponsive design and was the most preferred UI by participants. Finally, given the findings from experiments, as walking while being encumbered by carrying bags in public is common in everyday mobile use, the experiment in Chapter 6 explored the effect of encumbrance on gaze input. Specifically, gaze-only interaction using Dwell time, and multimodal interaction in which gaze is used for pointing and touch for selection. Touch input was also included to enable comparison of participants’ perception and preference for input modalities when walking and encumbered. This chapter aims to uncover the potential of employing gaze as an alternative modality to mitigate the impact of situational impairment on touch input on handheld mobile devices.
Omar Mohammed Namnakani (Thu,) studied this question.