Gaze Buttons Revisited.Looking for an Effective Way of "Clicking" without the Mouse

The fundamental problem for designing a gaze-based human-computer interaction is related to development of an effective method for activating graphical user interface elements by means of gaze only. Such a method should be easy for the user to apply, however at the same time, it requires eye movements that are clearly different from the natural behavior of the eye. We examined three methods of button activation by gaze, looking for the most effective way of gaze "clicking". These were: 1) the most standard method based on the use of dwell-time, 2) its modification based on detection of fixation located inside the buttons area and 3) and the most novel method based on gaze gestures consisting of movement into the button area and outward movement in the approximately opposite direction. We compared these gaze control methods under homogeneous conditions, which allows for a more reliable assessment of their relative usefulness. Two layouts of buttons were used: arranged on a grid, like on a telephone pad, and on a circle with an empty center. The experimental task was to enter a set of four-digit PINs using a set of gaze buttons corresponding to ten digits. A group of novices were instructed to use all the three methods and both button layouts (six experimental conditions). The activation methods were compared in terms of system usability, objectively measured by the PIN entry speed and the number of errors, as well as using a subjective SUS questionnaire. The system based on gaze gestures was worse in both measures; however, it had its followers. The method based on fixation detection instead of dwell-time did not significantly increase the entry speed due to the greater number of errors caused by non-intentional buttons activation. The circle layout turned out to be generally more convenient than the telephone pad layout.