Abstract: The eye tracking technique provides a new tool to interact with computers. Using an eye tracker, the gaze direction is detected and transformed to coordinates on the screen (i.e., gaze point). And when the gaze point fixates on a target beyond dwell time, the target is selected. Due to the fact that the dwell-based approach just uses eyes to select a target, it is very useful for disable people that do not have enough motor control abilities. However, the jittery nature of human eyes usually makes the gaze-controlled cursor unstable, which would impair the task performance and user experience. To counteract the instability of gaze cursor, we proposed the gaze-lock method to better handle normal and involuntary eye movements. Another problem to explore is whether to present real-time gaze cursor to the user, researchers have different views on this issue, but these views are based on the fact that the gaze point is equivalent to the gaze cursor. However, in our newly proposed gaze lock method, the gaze cursor is not completely equivalent to the real-time gaze point, so we explored the influence of hiding the real-time gaze point on eye-computer interaction performance on the basis of gaze lock. In a dwell-based eye typing experiment, we compared four cursor modes: the conventional gaze cursor, the no cursor, the gaze-lock cursor, and the gaze-lock partial cursor. The conventional gaze cursor is displayed as a hollow circle and updates its position based on the gaze point directly. The no cursor hides the visible cursor on the basis of the conventional cursor. In the gaze-lock cursor mode, two modifications are introduced to the conventional gaze cursor. First, when the gaze cursor stays in the target area beyond 150ms, it becomes locked at the target center. Second, when the gaze point falls out of the target area, the locked cursor will not be unlocked unless the gaze point leaves the target beyond 50ms. In the gaze-lock partial cursor mode, one difference from the gaze-lock cursor is that the displayed gaze cursor appears at the target center only when it is locked and disappears when it is unlocked. Twenty adults (including 3 females) were recruited, all of them were experienced computer users and familiar with Chinese Pinyin. The experimental task was to type the “#” symbol and subsequent Pinyin, ignoring case, as shown in Figure 2. The participant began a practice session, and subsequently performed a formal test session. Both the practice and test session included 4 blocks corresponding to 4 experimental conditions. For each participant, 80 trials were collected in total. Repeated-measure ANOVA was used to perform statistic analysis of typing speed and gaze shift time, and the non-parametric Aligned Rank Transform procedure was used to perform statistic analysis of KSPC, MSD error rate and gaze focus time. The results showed:(1) The gaze-lock method significantly improved the typing speed while keeping the error rate at very low level, the presence of real-time gaze cursor has no significant impact on input performance. (2) The gaze-lock method and presence of real-time gaze cursor significantly reduced the gaze shift time, but no significantly effect on the gaze focus time.(3) Moreover, the gaze-lock cursor was most preferred by the participants, followed by gaze-lock partial cursor, while no cursor and conventional gaze cursor are less preferred. In summary, our findings demonstrate that the gaze-lock method could improve the human performance and user experience in eye typing tasks, and have good potentials to be used in other dwell-based gaze interactions.

Key words: eye tracking, gaze input technology, gaze cursor mode, dwell-based paradigm

摘要： 显示与反馈是人机交互的重要设计要素。对于视线交互这种自然人机交互技术，如何呈现视线光标，尚存在争论。本研究选取基于凝视点击范式的文本输入任务，设计了2（有无视点锁定功能）*2（有无实时注视点）被试内实验，探讨光标显示模式对视线交互绩效与用户体验的影响。结果：视点锁定功能可以提升交互绩效和用户体验，有无实时注视点对交互绩效无显著影响；输入速度的提升主要在于该功能能够让被试的视线更快地转移到下一个目标上。研究结果为视线光标的设计与应用提供了指导。

关键词: 视线追踪, 视线输入技术, 光标显示模式, 凝视点击范式