Effectiveness of eye-gaze input method: Comparison of speed and accuracy among three eye-gaze input method

Abstract

Effectiveness of eye-gaze input methods was examined in click, drag, and menu selection tasks. In a click task, three eye-gaze methods were (c)-(i) eye-gaze input with fixation, (c)-(ii) eye-gaze input with pressing BS key, and (c)-(iii) eye-gaze input with voice (voice1). Method (d)-(i) eye-gaze input with pressing BS key and (d)-(ii) eye-gaze input with voice (voice1) were compared for the drag task. In the menus selection task, the performance was compared between Method (m)-(i) eye-gaze input with voice (voice1) and (m)-(ii) eye-gaze input with voice (voice2: uttering one of the following menu items: “save”, “print”, “cut”, “copy”, and “paste”). The pointing time in the click task increased according to the following order: (c)-(i) eye-gaze input with fixation, (c)-(ii) eye-gaze input with pressing BS key, and (c)-(iii) eye-gaze input with voice (voice1). The pointing accuracy of (c)-(i) was nearly equal to 100% and by far better than that of Method (c)-(ii) and (c)-(iii). Concerning the drag, Method (d)-(i) tended to be faster than Method (d)-(ii). However, the pointing accuracy of both methods was not satisfactory and ranged from 70% to 80%. This indicated that Method (d)-(i) and (d)-(ii) must be further improved when used for the drag task. The pointing time in the menu selection task did not differ significantly between Method (m)-(i) and (m)-(ii). The pointing accuracy of Method (m)-(ii) was by far higher than that of Method (m)-(i) when the target size was small. The larger target size tended to lead to faster and accurate pointing for all three tasks. It seems that the better pointing method differs according to the eye-gaze method. Other than the click task, the pointing accuracy was at most 90%. Therefore, future research should propose an effective method to increase the prediction accuracy for both drag and menu selection tasks.