GazeZoom: Exploration of Gaze-Assisted Multimodal Techniques for Panning and Zooming

Paper Title

GazeZoom: Exploration of Gaze-Assisted Multimodal Techniques for Panning and Zooming

Publication Info

• Topic area: Gaze-assisted multimodal interaction techniques for HMD-based zooming and panning.
• Keywords: Gaze interaction, head-mounted displays, zoom and pan, multimodal interfaces, eye tracking, hands-free interaction, VR usability, interaction techniques, accessibility, user experience.

Background and Problem

• Problem / challenge: Current HMD systems rely heavily on hand-based interactions for zooming and panning, which are inefficient, induce fatigue, and are unsuitable for scenarios where hands are occupied. Gaze-assisted techniques for zooming and panning remain underexplored.
• Significance: Efficient zooming and panning are critical for tasks like map exploration, design, and precise object manipulation in HMDs. Addressing these limitations can improve usability and accessibility in diverse applications.
• Motivation and related work: Prior work has shown the potential of gaze as a pointing mechanism and its integration with other inputs for selection tasks. However, existing techniques for zooming and panning often rely on fixed zoom centers or require both hands, limiting flexibility and usability in practical scenarios.

Solution

• Proposed approach: GazeZoom, a set of three gaze-assisted techniques for zooming and panning in HMDs, leveraging eye tracking for zoom center control and combining it with bimanual, unimanual, or head-based inputs.
• Novelty:
  1. Implementation of three distinct gaze-assisted techniques: BiGaze (bimanual), UniGaze (unimanual), and HeadGaze (hands-free).
  2. Systematic evaluation of these techniques against a baseline two-handed method.
  3. Design considerations for integrating gaze controls into HMD interactions.
  4. Demonstration of practical applicability in a simulated PCB soldering task.
• Procedure and key techniques:
  • BiGaze: Combines gaze for zoom center with two-handed gestures for zoom level and panning.
  • UniGaze: Uses gaze for zoom center and one-handed gestures for zoom and pan control.
  • HeadGaze: Integrates gaze for zoom center with head rotation for zoom control and eye gestures for panning.
  • Evaluation included task completion time, error rates, and subjective workload across 24 participants using a controlled alignment task.

Results

• Concrete findings:
  • BiGaze achieved the fastest task completion times and lowest error rates, with a 17%-36% speed improvement over the baseline.
  • UniGaze provided comparable precision with reduced physical effort.
  • HeadGaze enabled hands-free operation but required higher cognitive effort and had a steeper learning curve.
• Advantage over baselines:
  • All gaze-assisted techniques reduced manual panning and improved efficiency compared to the baseline.
  • BiGaze and UniGaze showed significantly lower NASA-TLX workload scores than the baseline.
• Experiments / evaluation:
  • Conducted with 24 participants performing 336 trials each in a controlled alignment task.
  • Measures included task completion time, error rates, cumulative zoom/pan changes, and subjective workload (NASA-TLX).
  • A follow-up demonstration study with 12 participants validated practical applicability in a PCB soldering scenario.
• Limitations and future work:
  • Current techniques rely on predefined parameters that may not suit all users.
  • Eye tracking reliability in commercial HMDs limits precision, especially for HeadGaze.
  • Future work should explore personalized parameter optimization, advanced sensor fusion, and longitudinal studies to assess learning effects.

Summary

This paper introduces GazeZoom, a set of gaze-assisted techniques for zooming and panning in HMDs, addressing limitations of existing hand-based methods. BiGaze, UniGaze, and HeadGaze leverage eye tracking for zoom center control, combined with bimanual, unimanual, or head-based inputs, respectively. Experimental results show that gaze-assisted techniques improve efficiency and usability, with BiGaze excelling in speed and precision, UniGaze balancing control and effort, and HeadGaze enabling hands-free operation. A follow-up demonstration in a PCB soldering task highlights their practical applicability. These findings contribute to the design of multimodal HMD interactions, particularly for scenarios requiring hands-free or accessible solutions.