Comparing Dwell time, Pursuits and Gaze Gestures for Gaze Interaction on Handheld Mobile Devices

Document Title

Comparing Dwell Time, Pursuits, and Gaze Gestures for Gaze Interaction on Handheld Mobile Devices

Document Information

• Subject Area: Human-Computer Interaction, Gaze Interaction Technology
• Keywords: Eye Tracking, Mobile Devices, Gaze Interaction, Smooth Pursuits, Human-Computer Interaction Technology

Research Background and Issues

• Problems and Challenges:

  • Various gaze-based interaction methods (e.g., dwell time, pursuits, and gaze gestures) have been proposed for mobile devices, but their performance in mobile scenarios remains underexplored.
  • Most gaze interaction studies have been conducted in static environments, making it difficult to directly apply the results to handheld mobile devices in motion.
  • In mobile scenarios, user instability during interaction (e.g., shaking while walking) significantly affects gaze estimation accuracy.

• Significance:

  • Gaze-based interaction can free users' hands, offering a novel interaction modality for mobile device users.
  • Understanding the performance of gaze interaction technologies on handheld mobile devices is crucial for advancing this technology.

• Research Motivation and Related Work:

  • Recent advancements in eye-tracking technology, coupled with improvements in mobile device front cameras and processing power, have made real-time gaze estimation feasible.
  • While the potential of methods like dwell time, pursuits, and gaze gestures on mobile devices has been studied, a systematic comparison in natural mobile scenarios is still lacking.

Solution

• Proposed Research Method:

  • Compare three common gaze interaction methods: Dwell Time, Smooth Pursuits, and Gaze Gestures.
  • Conduct experiments in both static and walking conditions, requiring participants to select varying numbers of targets (2, 4, 9, 12, 32) using their gaze.

• Research Innovations:

  • The first study to compare three gaze interaction techniques in mobile scenarios.
  • Covers both static and dynamic conditions, further exploring the impact of target quantity on interaction performance.
  • Designed experiments to validate the performance of these methods (e.g., selection time, error rate, cognitive load) and user preferences.

• Implementation Steps and Key Techniques:

  1. Design experimental tasks allowing participants to select targets on a smartphone screen using gaze.
  2. Evaluate the methods using performance metrics (selection time, error rate, timeout rate) and subjective metrics (NASA-TLX questionnaire).
  3. Perform statistical analyses to quantitatively and qualitatively compare the three methods.

Research Outcomes

• Specific Results:

  • Selection Time:
    • In static conditions, Pursuits was the fastest (1.36 seconds), followed by Dwell Time (2.33 seconds), with Gaze Gestures being the slowest (5.17 seconds).
    • In walking conditions, overall selection times were longer (Pursuits and Dwell Time outperformed Gaze Gestures).
  • Accuracy:
    • Gaze Gestures had higher accuracy than the other two methods, making it suitable for scenarios requiring high precision.
  • Cognitive Load:
    • Pursuits had the lowest cognitive load under static conditions, with users reporting less stress and fatigue.

• Advantages Over Existing Solutions:

  • Clearly reveals the performance differences of the three methods in mobile scenarios, providing guidance for gaze interaction design on mobile devices.
  • Offers performance rankings under different target quantities and mobility conditions, aiding in the selection of appropriate methods for practical applications.

• Experimental or Evaluation Results:

  • Pursuits performed better with fewer targets, while Dwell Time was preferred for larger target sets.
  • Gaze Gestures, though slower in multi-target environments, had extremely low error rates, making it suitable for high-precision scenarios.

• Limitations and Future Directions:

  • Limitations:
    • The experimental conditions were relatively ideal, such as obstacle-free indoor environments, which may underestimate real-world challenges.
    • The impact of target shape and size was not thoroughly investigated.
  • Future Directions:
    • Explore more robust gaze estimation algorithms to mitigate the effects of user posture and device vibrations on performance.
    • Investigate more natural and intuitive interaction mechanisms to enhance user experience.