Depth Perception in Virtual Environments

@ York University

This project was for the requirements of a PhD in Psychology (Brain, Behaviour & Cognitive Science) at York University. The goal of these three studies was to investigate how users perceive and interact with objects in virtual reality environments relative to natural viewing environments.

Background & Goals

The goal of this series of studies was to understand how factors such as viewing distance, cue conflicts, and interactability contribute to depth distortions in virtual environments. To do so, I conducted a series of studies in which users estimated depth under a range of viewing conditions and cue combinations.

Research Goals

Measure depth distortions of virtual relative to physical objects
Assess limitations of virtual environments when integrating motion and binocular cues
Evaluate if reaching towards virtual objects provides additional cues to scale depth perception

Methods

Three experimental studies were developed using Unity and/or Python to render objects in virtual reality environments in multiple display systems (including head-mounted displays and stereoscopes). Blender was used to model all virtual objects and 3D print perceptually matched physical models that were displayed in a physical test environment.

All data were analyzed using advanced statistical techniques, such as psychophysical modelling, signal processing, multiple linear regression, mixed models, logistic regression, Bayesian statistics, and bootstrapping using R.

Figure 1: Sample of physical object

Crucial Insights

Participants underestimated distance and depth of virtual objects while estimates of physical objects were accurate and exhibited depth constancy. This failure was due in part to the presence of vergence-accommodation conflict in virtual displays.

Participants relied more heavily on binocular depth cues than motion cues. Interacting with virtual objects did not enhance performance on tasks dependent on distance perception. To fully understand how binocular depth perception is used to interact with objects in the real world, it is important to assess these cues in a rich, full-cue natural scenes.

Research Impact

Strategic Impact

Published studies contributed important considerations of (1) context of the depth cues in the scene, (2) participant's prior experience, and (3) how that relates to natural viewing environments to the scientific literature regarding depth perception
Improved ecological validity of perceptual research by evaluating 3D perception using naturalistic objects, especially given the majority of other studies focus only on computerized displays

My Learnings

Conceptualizing, designing, programming, and executing long-term end-to-end research projects that answer ambiguous research questions using creative and innovative problem solving
Experience using Unity, Python, and R to develop complex perceptual experiments and conduct advanced data analyses
3D modelling experience using Blender to create naturalistic virtual objects and scenes