Stretch Gaze Targets Out: Experimenting with Target Sizes for Gaze-Enabled Interfaces on Mobile Devices

Research Background and Issues

• What problems or challenges did the authors identify?
  Previous studies have explored gaze interaction techniques on mobile devices, such as Dwell Time and Pursuits, but these techniques have not sufficiently considered the impact of changes in distance between the user and the screen on gaze selection accuracy and design optimization. Additionally, vertical gaze tracking accuracy is relatively low, and the correlation between screen position and target size remains unclear.

• Why is this issue important?
  The usage scenarios of mobile devices are highly variable, such as users placing devices arbitrarily or interacting in dynamic environments (e.g., standing or walking), which affects gaze tracking accuracy. Optimizing gaze interface design is crucial for enhancing user experience and reducing visual fatigue, especially in cases where users face physical constraints (e.g., one-handed use) or when touch interaction is inconvenient.

• Research Motivation and Related Work
  The authors’ work addresses existing challenges in gaze interaction, particularly the impact of target size on tracking accuracy. Considering screen regions, target sizes, and distances can help developers optimize gaze interaction interfaces while addressing design deficiencies in previous studies that did not fully account for mobile device scenarios.

Solution

• What methods or solutions did the authors propose?
  The authors designed an experiment to measure the effects of different target sizes (visual angles), screen regions, and user-device distances on gaze tracking accuracy, deviation, and visual fatigue, providing developers with guidelines for target size design.

• What are the innovative aspects of this solution?

  1. Separating the discussion of target dimensions (visual angle) from tracking region size, exploring the impact of expanding the tracking region (rather than visual size) on tracking performance.
  2. Dynamically adjusting target sizes for mobile device usage scenarios to accommodate changes in user-screen distance.
  3. Conducting detailed tests on horizontal and vertical tracking accuracy differences, revealing the phenomenon of more precise horizontal gaze tracking.

• What are the implementation steps and key technologies used?

  1. Using the EyeDid eye-tracking library on an iPhone X device to record gaze data in real time.
  2. Conducting experiments in a controlled environment, testing five distance ranges (25–49 cm), four target sizes (2°–5° visual angles), and three screen regions (top, middle, bottom).
  3. Analyzing gaze data for accuracy, precision, and offset errors (horizontal and vertical directions).
  4. Investigating the impact of expanding the tracking region, particularly vertical expansion.

Research Outcomes

• What specific outcomes were achieved?

  1. Optimal Target Size Recommendation: A visual angle of 4° for the target was identified as the "optimal balance point," maintaining over 70% gaze tracking accuracy across most distances.
  2. Horizontal Outperforms Vertical: The experiments showed that horizontal tracking accuracy and precision were significantly better than vertical tracking, likely due to the "horizontal-vertical anisotropy" phenomenon in human visual perception.
  3. Effect of Tracking Region Expansion: Expanding the vertical tracking region significantly improved tracking accuracy, particularly for smaller target sizes.

• What advantages does it have compared to existing solutions?

  1. More accurately measured the impact of dynamic changes in user-screen distance on gaze interaction.
  2. Provided practical optimization suggestions tailored to mobile device design, rather than fixed scenarios (e.g., desktop eye trackers).
  3. Addressed the issue of lower vertical tracking accuracy by expanding the vertical tracking range.

• What were the experimental or evaluation results?
  Experimental results indicated:

  • A visual angle of 4° for gaze targets achieved over 70% tracking accuracy across all screen regions and distance ranges.
  • Smaller targets (e.g., 2° visual angle) significantly improved gaze tracking performance after expanding the vertical tracking region.
  • Vertical tracking offset was significantly higher than horizontal offset, especially at the top and bottom regions of the screen, where users’ gaze points tended to drift toward the screen center.

• Limitations and Future Directions
  Limitations:

  • The experiments relied on specific mobile devices (iPhone X) and the EyeDid SDK, limiting generalizability across different devices and tracking technologies.
  • The screen position was fixed at a 60° angle, not simulating more complex user usage angles.

  Future Directions:

  • Investigating gaze interaction performance in real dynamic environments (e.g., walking or seated positions).
  • Integrating machine learning and sensor data (e.g., automatic detection of user holding posture) to dynamically adjust target design.
  • Researching techniques to reduce gaze errors at screen edges, such as intelligent correction or dynamic adjustment of tracking regions.