MiniMates: Miniature Avatars for AR Remote Meetings within Limited Physical Spaces

Research Background and Issues

• Identified Problems or Challenges:
  The authors point out that current video conferencing tools (e.g., Zoom) have limitations in facilitating natural communication, particularly in directional communication and simultaneous group discussions. These tools are ineffective in conveying non-verbal communication cues such as body orientation, head movements, or eye contact. Additionally, traditional video conferencing relies on a single audio channel, restricting natural side conversations. While some virtual reality (VR)-based approaches address certain issues, they isolate users in virtual environments, lacking awareness of the surrounding physical environment.

• Significance:
  Non-verbal cues (e.g., body orientation, eye contact) are crucial for understanding context and enabling natural communication. Furthermore, with the increasing demand for online meetings, enhancing remote collaboration efficiency in confined spaces has become increasingly important.

• Research Motivation and Related Work:
  This study is inspired by related works, including research on directional communication in video conferencing, virtual avatar positioning methods in AR and VR, and experiments using miniature avatars. The authors aim to leverage AR technology to address the limitations of traditional video conferencing tools and immersive VR methods, using innovative miniature avatar designs to meet user needs in space-constrained environments.

Solution

• Proposed Solution:
  The authors propose a novel AR-based remote conferencing system called MiniMates, which employs miniature 3D avatars. MiniMates effectively places multiple remote participants within the user's limited physical space, minimizing interference between virtual avatars and the physical environment, while optimizing user interaction through directional communication algorithms.

• Innovations:

  1. Miniature Avatars: Compared to life-sized virtual avatars, miniature avatars reduce spatial occupancy, making them suitable for small desks or home environments.
  2. Directional Gaze and Posture Redirection Algorithm: Ensures that all participants can use non-verbal cues to understand who is communicating with whom, maintaining the natural flow of conversations.
  3. Spatial Layout Formation Algorithm: Provides consistent avatar arrangements to address variations in local space size and layout.

• Implementation Steps and Key Technologies:

  1. Avatar Miniaturization and Layout: Using computer vision algorithms, avatars are scaled down to approximately 20% of their original size and arranged in a semicircular layout on the desk to fit confined spaces.
  2. Redirection Algorithm: Dynamically adjusts virtual avatars based on users' head and body orientations to ensure consistent interaction experiences from all users' perspectives.
  3. Layered Audio Processing: Synchronizes sound sources with avatar positions to enhance the natural spatial audio experience.
  4. Additional Functional Extensions: Features such as cursor interaction allow users to move avatars during screen sharing to enhance collaboration.

Research Outcomes

• Specific Results:

  • MiniMates effectively enables natural non-verbal communication, allowing users to interact with others directly through body orientation and head direction.
  • The system significantly enhances users' sense of presence, particularly in scenarios involving multiple participants and simultaneous group discussions.

• Advantages Compared to Existing Solutions:

  • Compared to traditional 2D video conferencing, MiniMates greatly improves users' sense of co-presence and social engagement.
  • Compared to fully immersive VR conferencing, MiniMates maintains users' awareness of their physical surroundings, making it more suitable for everyday work scenarios.

• Experimental and Evaluation Results:

  • A user study involving 24 participants revealed that MiniMates scored significantly higher than traditional video conferencing across multiple dimensions of the Telepresence Inventory (TPI), particularly in terms of user appeal and immersion.
  • In the "First Contact Task," which allowed simultaneous side conversations, users efficiently engaged in parallel dialogues—something deemed infeasible with traditional video conferencing tools.
  • In non-verbal directional tasks, the error rate was extremely low (only 2 errors), validating the effectiveness of the system's directional communication functionality.

• Limitations and Future Directions:

  • Current Limitations: Miniature avatars may struggle to accurately convey facial expressions. Additionally, current testing is limited to five participants, and the system has not yet been optimized for larger-scale meetings.
  • Future Directions:
    1. Improve facial expression recognition and rendering, exploring the potential of high-definition hardware to enhance miniature avatar representation.
    2. Extend to dynamic scenarios, such as users moving within virtual spaces or interacting across groups.
    3. Develop smarter volume adjustment algorithms to reduce interference from additional conversations during group sessions.
    4. Test other task types (e.g., lecture-style meetings) and more complex collaborative activities.