SMOOVS: Towards calibration-free text entry by gaze using smooth pursuit movements

Abstract

Gaze-based text spellers have proved useful for people with severe motor diseases, but lack acceptance in general human-computer interaction. In order to use gaze spellers for public displays, they need to be robust and provide an intuitive interaction concept. However, traditional dwell- and blink-based systems need accurate calibration which contradicts fast and intuitive interaction. We developed the first gaze speller explicitly utilizing smooth pursuit eye movements and their particular characteristics. The speller achieves sufficient accuracy with a one-point calibration and does not require extensive training. Its interface consists of character elements which move apart from each other in two stages. As each element has a unique track, gaze following this track can be detected by an algorithm that does not rely on the exact gaze coordinates and compensates latency-based artefacts. In a user study, 24 participants tested four speed-levels of moving elements to determine an optimal interaction speed. At 300 px/s users showed highest overall performance of 3.34 WPM (without training). Subjective ratings support the finding that this pace is superior.