[Paper Review] AvatarGo: Plug and Play self-avatars for VR

José Luis Pontón, Eva Monclús+1 more|arXiv (Cornell University)|Jan 1, 2022

Virtual Reality Applications and Impacts7 citations

TL;DR

AvatarGo presents a plug-and-play system for creating personalized self-avatars in VR using low-cost trackers (HMD, 2 controllers, 3 trackers). It computes user-specific offset values between body trackers and avatar joints via a simple walk-in calibration, significantly improving pose accuracy and embodiment. User studies show a statistically significant increase in sense of embodiment (p < .001, d = 0.60) when using exact offsets versus fixed offsets.

ABSTRACT

The use of self-avatars in a VR application can enhance presence and embodiment which leads to a better user experience. In collaborative VR it also facilitates non-verbal communication. Currently it is possible to track a few body parts with cheap trackers and then apply IK methods to animate a character. However, the correspondence between trackers and avatar joints is typically fixed ad-hoc, which is enough to animate the avatar, but causes noticeable mismatches between the user's body pose and the avatar. In this paper we present a fast and easy to set up system to compute exact offset values, unique for each user, which leads to improvements in avatar movement. Our user study shows that the Sense of Embodiment increased significantly when using exact offsets as opposed to fixed ones. We also allowed the users to see a semitransparent avatar overlaid with their real body to objectively evaluate the quality of the avatar movement with our technique.

Motivation & Objective

• To address the lack of accurate, user-specific pose matching between real body movements and virtual avatars in VR.
• To reduce setup complexity and eliminate manual tweaking of tracker-to-joint offsets in avatar animation pipelines.
• To enhance the sense of embodiment (SoE) by minimizing visual-motor mismatches between user and avatar.
• To provide a plug-and-play solution compatible with any VR application using standard hardware and Unity.
• To objectively evaluate avatar fidelity using a semitransparent overlaid avatar mode with pass-through HMD cameras.

Proposed method

• Uses a T-pose calibration step where users stand in place and align trackers (on feet and lower back) with a virtual avatar rendered in a mirror.
• Automatically assigns tracker roles (root, feet) based on spatial clustering of tracker positions on a fitted plane.
• Computes exact 3D position and rotation offsets between each tracker and corresponding avatar joint (root, feet, head, back, hands) at calibration time.
• Applies gradient descent to minimize the distance between finger joints and controller capsules, enabling natural hand positioning on controllers.
• Uses signed distance functions (SDFs) to model controller geometry and optimize finger joint positions via numerical gradient updates.
• Stores computed offsets and uses them in inverse kinematics (IK) pipelines to drive avatar animation with high fidelity.

Experimental results

Research questions

• RQ1 Can a simple, automated calibration process significantly improve the accuracy of avatar movement compared to fixed offset assumptions?
• RQ2 Does computing user-specific offsets lead to a measurable increase in the sense of embodiment (SoE) in VR?
• RQ3 How does visual feedback (via pass-through camera overlay) affect users’ perception of avatar fidelity and embodiment?
• RQ4 To what extent does the system improve hand positioning on controllers, enhancing visual-haptic alignment?
• RQ5 Can the system be deployed as a plug-and-play solution across diverse VR applications without complex integration?

Key findings

• Users reported a statistically significant increase in sense of embodiment (SoE) when using exact offsets (p < .001, Cohen’s d = 0.60) compared to fixed offsets.
• The sense of embodiment was significantly higher for exact offsets in both virtual avatar (VA) and overlaid avatar (OA) conditions (p < .001, d = 1.00 and d = 1.03 respectively).
• The drop in SoE when viewing the overlaid avatar was more pronounced for fixed-offset avatars, indicating greater visual-motor mismatch.
• Median SoE scores were 6 (IQR = 1) for exact offsets with virtual avatar only, indicating strong user agreement on embodiment.
• Median scores for sense of agency (SoA) and sense of ownership (SoO) were both 6 (IQR = 1), indicating high user confidence in controlling and identifying with the avatar.
• The system successfully computed plausible hand poses over controllers via SDF-based optimization, improving visual-haptic alignment.