DialPlates: Enabling Pursuits-based User Interfaces with Large Target Numbers

In this paper we introduce a novel approach for smooth pursuits eye movement detection and demonstrate that it allows up to 160 targets to be distinguished. With this work we advance the well-established smooth pursuits technique, which allows gaze interaction without calibration. The approach is valuable for researchers and practitioners, since it enables novel user interfaces and applications to be created that employ a large number of targets, for example, a pursuits-based keyboard or a smart home where many different objects can be controlled using gaze. We present findings from two studies. In particular, we compare our novel detection algorithm based on linear regression with the correlation method. We quantify its accuracy for around 20 targets on a single circle and up to 160 targets on multiple circles. Finally, we implemented a pursuits-based keyboard app with 108 targets as proof-of-concept.