Hochschulschrift:
Dissertation, Universität des Saarlandes, 2019
Anmerkungen:
Beschreibung:
Recently, gaze-based interfaces in stationary settings became a valid option, as remote eye trackers are easily available at a low price. However, making gaze-based interfaces ubiquitous remains an open challenge that cannot be resolved using remote approaches. The miniaturization in camera technology allows for gaze estimation with head-mounted devices, which are similar to a pair of glasses. A crucial step towards gaze estimation using such devices is calibration. Although the development is mainly technology driven, a hypothetical fully calibrated system in which all parameters are known and valid is affected by calibration drift. In addition, attributes like minimal intrusiveness and obstruction, easy and flexible setup and high accuracy are not present in the latest commercial devices. Hence their applicability to spontaneous interaction or long-lasting research experiments is questionable.In this thesis we enable spontaneous, calibration-free and accurate mobile gaze estimation. We contribute by investigating the following three areas: Efficient Long-Term Usage of a head-mounted eye tracker; Location, Orientation & Target Independent head mounted eye tracking; and Mobile & Accurate Calibration-Free gaze estimation. Through the elaboration of the theoretical principles, we investigate novel concepts for each aspect; these are implemented and evaluated, converging into a fi nal device, to show how to overcome current limitations.