VR Physicality Errors & Human Perception

• Errors that arise due to a mismatch in motion dynamics and visualized movements in VR are termed "physicality errors."
• Unlike simple physical errors like footskate, physicality errors involve plausible motions but with inconsistent dynamics.
• Physicality errors are inevitable when avatars don't match proportions or lift virtual objects with unexpected weight.
• A study explores people's sensitivity to physicality errors using a dumbbell lift exercise to understand when these errors are noticeable.
• Kinematic (motion) information primarily impacts effort perception; conversely, visual size of lifted object strongly influences perceived weight.
• Both can lead to perceptual mismatches, reducing perceived naturalness.
• Small errors might go unnoticed, but large errors reduce naturalness significantly
• These findings will inform animation algorithm requirements.

Human Perception in VR

• An aim is for embodied virtual reality and metaverse scenarios, allowing motion-driven avatars without appearance restrictions.
• Believable avatars for virtual companions are important for a sense of co-presence, which requires understanding and addressing errors.
• Research is aimed at impact of physicality errors on user experience from observing avatars, using dumbbell lifting as a model
• The research aims to understand perceptual impacts of kinematic and visual signals

Key questions asked

• Baseline: Accuracy in perceiving effort and inferring weight from motion kinematics on a VR character.
• Body Shape: Sensitivity to motion dynamic errors when avatars have different proportions or mass.
• Dumbbell Size: Sensitivity to avatars moving different masses than what they actually moved.
• Strain Deformations: How facial strain shifts the perception of effort and inference of weight in lifts.
• Discrimination: Ability to distinguish between zero-weight lifts and accurate lifts with varying weight dumbbells.

Study Design and Findings

• Participants watch animations of avatars lifting dumbbells, estimating effort and weight, and rating naturalness.
• Avatars and their bodies are matched with the lifters, with blendshape models for realistic muscle deformations.
• Kinematic data primarily drives judgments of effort and weight.
• Scaling is influenced by the size of the avatar body.
• Visual information of weight shifts effort estimates.
• Body and dumbbell sizes affect weight estimates more than effort estimates, which creates discordant signals.
• Muscle strain impacts effort and weight estimates when dumbbells aren't visible.
• Strain signals make incorrect motion kinematics less noticeable with appropriate strain for the dumbbell weight.

Background

• Research on weight perception spans computer graphics, VR, perception, and psychophysics.
• Perception of animated character motion has been studied, for example child versus adult motion. which includes virtual interactions.
• Expressive animation tools like Laban Movement Analysis address these perceptual errors.
• Perception of weight in VR is extensively studied.
• Pseudo-haptics can manipulate visual information to create haptic sensations
• Kenny et al. measured sensitivity to mismatches between avatar body size and motion and found limited mismatch detection and naturalness.
• The experiment did not visualize the objects that the avatars interacted with, affecting the interpretation.

Human Motion Dynamic

• Research uses point-light displays to estimate weight of lifted objects or dumbbells with widely varying estimation accuracy.
• Factors that improved estimates include showing a reference lift and gaining haptic experience.
• Full-video resulted in highest accuracy compared to point-light displays.
• Both motion and human form are key in judging actions with muscle strain cues
• Object velocity decreases correlated with object weight with other kinematic changes observed such as Dwell time and hip angle.
• Kinematic patterns influence weight perceptions
• Simulating neck strain cue and considering the motor cortex helped.
• Size of the box does not confound weight perception, and visual cues integrated manipulated forces during precision grip.
• Integration of Effort and Visuals:
• If lifter's size and weight are known beforehand, weight and effort estimates become comparable.
• Observers more attuned to effort than weight, according to studies.

Psychophysics and Perception

• There is a power is low relationship between perception and physical quantity in psychophysics with people feeling that an object with 70% weight is half heavy.
• Size-weight illusion (SWI) is when smaller objects are judged heavier
• Factors such is objects material effect the perception of heaviness.

Experiments in VR

• Series of experiments explore signals of kinematic, avatar object representation, and muscle deformations. The design included survey in VR followed by clip ratings.
• Perceived effort was rated from 0-100% and motion naturalness was rated on a 7 point likelihood Scale and resulted in a brief Exit interview debrief. Unity was used to develop the Quest 2 apparatus experiment where the VE a normal room.
• Interactions in the VE used Oculus controller. Actors are acquired via online advertisement.
• Their motion sources for the study was a series of dumbbell curl lifts and then selected on two lifter sets on a "strong" and "average" scale; maximum lift capability.

Visual & Kinematic Factors and Methodology

• Motion of lifters captured with mocap.
• Markers are solved with skeletons on a custom solver; virtual replicas duplicated scaling volume assuming pixel density.
• Avatar needed to vary in a body shape providing plausible muscle deformations.
• Blend shapes control deformations in high-level strain rather than tune; the deformations are artist-driven.
• Linear mixed effect tests evaluate significance within the models and Wald chisquare tests with the participant tested as random, as their pool is just a sample of the population.

Body Shape Analysis

• Four different avatars are driven by the recorded motion of both average and strong lifters to allow body type, motion, and displayed weight combinations.

Results indicate

• Kinematic information is largely used to judge effort/weight
• Visual size impacts weight estimates more than effort estimates, causing estimation discrepancies.
• Naturalness ratings decline when major mismatches occur.

Perception of weight and effort are essential for animation

• Visualized virtual environments must correctly represent body size and dynamic motion.
• Experiment 1 aims to establish how accurately people can perceive the effort and infer the weight
• dumbbell lifting motions with varying weights are shown
• Participants estimate effort and weight
• clips grouped by lifter's maximum lift.
• A high correlation suggests ability for people to infer weight from lifters' kinematics.

Visual effects of the results

• Correlations stronger for effort, stronger lifters, higher effort, with effort directly observed and weight needing inferred.
• Curvilinear relationship between actual and estimate.
• The model was fitted and the Weight levels differ across the lifters.
• Linear mix effect indicates how people judge on a model.
• An attempt to normalize results to determine the relation between effort and rating estimate indicates strong and clear correlation in perception of how people make and estimate from motion.
• Weight and kinematics
• Kinematics accurately conveys effort and weight
• Sensitivities to kinematic variations is reduced when there is a low weight
• People were accurate when lifter was larger unlike previous tests.

Body Shape Perception

• The goal was to understand avatar, impact on weight, effort estimation using a VR.
• The size on VR size as a use factor
• Realistically the dumble size on each of the four lifters' bodies with the motion-visual signal that can be inconsistent. Impact on the various perceptions.
• There was a minor impact on perceived effort overall. However, there was still a significant impact on Body.
• Effect in a weight the fact that the body suggests people assume the larger avatars lifted. Larger bodies can lift objects heavier given equal effort based on observation. 11 average on Strong avatar 11% impact.

Dumbbell Effect

• Impact as a cause to change to naturalness from the naturalness perspective.
• All were impactful to body shape.
• Dumbbells lower levels as if participants in lift is believed.

Further study design and experiment

• Kinematic effect on weight not impacted by dumbbell visualization where you increase correlation using Fisher transformation.
• Smaller impact on avatar indicates more sensitivity in that people perceive the correct lift unlike avatar's body.
• Impact of different mass.
• Display on avatar body.

A main take away is the relationship between estimated effort and visual

• Clear effect in how to the weights on that impression and naturalness.
• These can be less natural like with light weight were made on the avatar.

How Muscle strains and forms the Perception.

• A main investigation is relative impact by having different muscle strains and different muscle categories.
• How is applied to show the manipulation check which gives them the best.

Takeaways:

• In increasing Effort and order increases by a body, but with separated results, those related head impact with head fully clear as large as a large indicator.
• Weight has impacted large impact on the full deformation by full effect with limited on the impacts lowest left.
• This indicates as people a muscle to weight

When discrimination is correct and is the first two understand

• In comparison to having the correct
• There is a challenge using hand tracing to see, so an impact.
• The more is in detail the person would perform and then. In the experiment and two.
• Is the correct visualization like from where that which way around.
• This is a bit experiment by making their zero more obvious.

Results

• People performed more on chance level with full lift discrimination over chance level. This translates into adjustments to VR where they are not lifting very light objects. Even though we are likely required. Also is known that which one can indicate a weight will depend on how good something is based.
• Deformities and Kinematics: Even then they can try and test by being a simple line they is how to apply more with that to their is how does effect or effort of both.
• At strain level they are to indicate but is.

Visual size indicators and kinematics

• It is like being the weight of an is not how weight estimate but it is only to be for lift
• They can have a weight that can lead to be over estimates
• These are small and can try and adjust.
• In muscle deformation there has been a great improvement and by all is made for in the deformation.

Limitation

• It is because they are limited in how that.

In conclusion

This is made of which is how can weight and of that. It is clear that what is not. There are limited by using. It can increase even that has to have the strain of that.