Snap, Pursuit and Gain: Virtual Reality Viewport Control by Gaze

Title of the Paper

Snap, Pursuit and Gain: Virtual Reality Viewport Control by Gaze

Paper Information

• Research Domain: Virtual Reality (VR), User Interaction, Viewport Control Techniques
• Keywords: Eye Tracking, Gaze Interaction, Eye-Head Coordination, Viewport Control, User Study, Eye-Based Interaction, Virtual Reality

Research Background and Issues

1. Problems and Challenges:

   • Virtual Reality (VR) typically relies on head movement to control the viewport (via head-mounted displays, HMDs). However, in certain situations, such as when users cannot move their heads or their body posture is restricted (e.g., lying down or due to injury), this approach becomes limited.
   • Traditional viewport control methods also include controller-based manipulation, but this approach cannot meet hands-free interaction requirements.

2. Significance:

   • Providing viewport control methods that do not rely on head or hand movements is crucial for enhancing the accessibility and universality of VR systems.
   • Improving user experience in constrained spaces (e.g., public transportation) or environments requiring more stable interaction.

3. Research Motivation:

   • Exploring alternative interaction methods, such as using eye tracking for viewport control, allows users to reduce physical movement demands and optimize interaction efficiency.

4. Related Work:

   • VR navigation techniques and viewport control have been widely studied, including hands-free solutions using controllers, head amplification, and voice control. However, no existing literature has thoroughly investigated gaze-based viewport control.

Solution

1. Method Overview:

   • Three gaze-based viewport control techniques are proposed: “Dwell Snap,” “Gaze Gain,” and “Gaze Pursuit.”
   • These methods utilize different types of eye movements (fixation, saccades, smooth pursuit) to achieve 360° viewport control and integrate users’ natural eye-head coordination responses.

2. Innovations:

   • First-time exploration of gaze-based viewport control, relying entirely on eye movements without requiring head movement.
   • Mapping natural eye-head coordination mechanisms to technical design, enabling dynamic adjustments for complex scenarios.
   • Proposing three interaction modes tailored to different situational needs, incorporating characteristics of the human visual system.

3. Key Techniques and Steps:

   • Dwell Snap: Based on eye fixation, triggering a jump after the gaze remains at a certain angle for a specific duration.
     • Trigger mechanism angle threshold: 25°, initial dwell time: 400 milliseconds, later reduced to 200 milliseconds.
   • Gaze Gain: Amplifies the combined output of eye and head movements to enhance viewport rotation.
     • This technique uses dynamic speed adjustment to mitigate dizziness and unintended jumps caused by rapid scene updates.
   • Gaze Pursuit: Smoothly tracks the user's gaze direction until the fixation point aligns with the viewport center.
     • Enables dynamic viewport adjustment based on gaze, supporting smoother target alignment operations.

Research Findings

1. Main Results:

   • The three gaze-based techniques demonstrated comparable task completion efficiency and viewport alignment accuracy to traditional techniques (e.g., controller-based jumps, head amplification).
   • Experiments showed the effectiveness of Dwell Snap, Gaze Gain, and Gaze Pursuit in various viewport control tasks, supporting hands-free control needs and applications in specific scenarios.

2. Comparison with Existing Solutions:

   • Overall group performance indicated no significant differences between gaze-based techniques and traditional controller or head movement methods in terms of error rate, task time, user load, and simulated dizziness.
   • Unlike traditional methods, gaze-based control reduced reliance on head and hand movements, making it more suitable for constrained scenarios or quiet interaction needs (e.g., using VR on public transportation or while lying down).

3. Experimental Evaluation Results:

   • In precise control tasks, certain techniques (e.g., Gaze Gain) showed slightly higher error rates, but no significant differences were observed in relaxed scenarios.
   • No significant differences in cumulative head and eye movement amplitudes across techniques, though individual differences in eye-head coordination tendencies were observed during gaze-based interaction.
   • Subjective user load (NASA TLX scale) indicated slightly increased physical load for Dwell Snap, but overall remained manageable.

4. Limitations and Future Directions:

   • Limitations:
     • Tasks were abstract experimental scenarios, lacking evaluation in complex backgrounds or multi-target interactions.
     • Participants were seated, with no extension to more constrained postures (e.g., lying down or standing).
     • Individual differences (eye-head coordination tendencies) masked performance differences between techniques.
   • Suggested Future Work:
     • Validate the performance and adaptability of these techniques in real-world task contexts.
     • Explore improvements to accommodate individual diversity (e.g., eye-head coordination tendencies).
     • Extend the application of these techniques to Augmented Reality (AR) and complex scenarios (e.g., 360° video playback, immersive education).