immersivePOV: Filming How-To Videos with a Head-Mounted 360° Action Camera

Title of the Paper

immersivePOV: Filming How-To Videos with a Head-Mounted 360° Action Camera

Paper Information

• Authors: Kevin Huang, Jiannan Li, Maurício Sousa, Tovi Grossman
• Institutions: University of Toronto, Simon Fraser University
• Conference: CHI Conference on Human Factors in Computing Systems (CHI ’22)
• Year: 2022
• Research Area: Human-Computer Interaction, Online Learning and Instructional Video Design
• Keywords: how-to video, online learning, YouTube, 360° video, POV

Research Background and Problem

• Identified Issues or Challenges:

  • A large number of online instructional videos still use third-person perspective (3pp), which may hinder viewers' learning due to the "perspective effect."
  • First-person perspective (1pp) videos lack sufficient immersion and fail to fully utilize modern technologies.
  • There is no clear framework guiding video design choices, with creators relying more on intuition.

• Significance:

  • Effectively designing instructional videos, particularly in presenting perspectives that are easier for users to understand, is a crucial direction for improving the online learning experience.
  • Enhancing learning for practical tasks not only supports viewers but also simplifies the video creation process.

• Motivation and Related Work:

  • While existing research suggests that 1pp may be more effective than 3pp, most implementations use "quasi-POV" perspectives (e.g., shoulder or close-up shots), which lack immersion.
  • The potential of 360° VR videos in education has been preliminarily explored, but their practical application and quality still need improvement.

Solution

• Proposed Method or Solution:

  • The authors propose an immersive first-person perspective (immersivePOV) method for filming instructional videos using a head-mounted 360° action camera.
  • Videos can be viewed through VR headsets, offering eye-level perspectives and three degrees of freedom (3DoF) for interactive experiences.

• Innovative Features:

  • Introduces a "what-you-see-is-what-you-(would)-get" (WYSIWYWG) high-immersion video experience.
  • Combines the advantages of 360° VR and POV technologies, providing viewers with enhanced spatial and contextual immersion.
  • Simplifies production complexity, enabling point-and-shoot functionality without the need for editing or multi-camera setups.

• Implementation Steps and Key Technologies:

  • Video Production:
    • Uses a GoPro MAX 360° camera mounted on a helmet to ensure eye-level perspective.
    • Generates standard POV video versions using the Overcapture tool in the GoPro app.
    • Films continuous long shots in 360° VR mode without switching perspectives.
  • User Experiment:
    • Compares three video conditions (3pp, traditional POV, and immersivePOV) to observe viewers' task performance and experience.
    • Videos are played via YouTube or VR devices.

Research Findings

• Specific Results:

  • Experiments show that immersivePOV videos significantly reduce viewers' cognitive load, improve task completion efficiency, and increase user preference.
  • Compared to traditional POV, immersivePOV demonstrates higher learning effectiveness due to the "immersion effect."

• Advantages Over Existing Solutions:

  • Compared to 3pp, immersivePOV avoids learning difficulties caused by spatial transitions.
  • Compared to traditional POV, immersivePOV offers stronger immersion, helping viewers feel as though they are performing the task themselves.

• Experiment or Evaluation Results:

  • Task Learning: Participants in the immersivePOV group completed tasks in the shortest time, primarily exhibiting continuous viewing behavior rather than skipping or rewinding.
  • Cognitive Load: Participants in the immersivePOV group reported the lowest mental burden while watching and attempting tasks.
  • User Feedback: Participants widely acknowledged the WYSIWYWG feature of immersivePOV and found it very easy to follow.

• Limitations and Future Directions:

  • Task Selection: The study used relatively simple tasks (e.g., folding shirts and tying shoelaces), and the effectiveness for complex and dynamic tasks remains to be verified.
  • Usage Scenario Limitations: The study only examined asynchronous video viewing behavior, without exploring real-time interaction (e.g., AR combined with immersive instructional videos).
  • Equipment Requirements: immersivePOV relies on VR devices, which may introduce sample bias due to device ownership.
  • Future Work:
    • Explore learning effects for more complex tasks, including long-duration dynamic tasks and interactive tasks.
    • Investigate long-term memory retention of task learning.
    • Further implement the approach in educational settings, such as engineering and medical skill training.

Through this study, the authors provide a new perspective for instructional video design and highlight the potential of head-mounted 360° action cameras as tools for video production and learning.