Physiological and Psychological Responses to Urban Greenery - Immersive Experiment Using Virtual Reality and Multimodal Measurement

NYU Center for Urban Science + Progress & Global Public Health, University of Florida, Fordham University
(Mentor: Prof. Anton Rozhkov, Prof. Prince Amegbor, Prof. Zhaoxi Zhang, Prof.H. Shellae Versey)

American Association of Geographers (AAG) Annual Meeting (2026)

News🚩: Our IRB is approved! Click here for more information and participate!

Understanding how people experience urban greenery requires more than correlations between “green metrics” and health outcomes. Real streets are full of confounding factors—traffic noise, weather, building form, social context—which makes it hard to isolate which visual elements actually shape comfort and stress. In this project, I shifted from observational studies to controlled, computational experimentation, using VR and biosensing to examine the micro-level cognitive mechanisms behind environmental appraisal.

We built a VR experimental framework that uses 360° panoramas of New York City streetscapes, computes Green View Index (GVI), and then generates controlled stimuli through AI-based visual manipulation. By combining subjective ratings with physiological signals (EEG/EDA), we can probe how incremental changes in street greenery influence attention, arousal, and perceived restorativeness.

Experimental Framework Development: Designed a controlled VR setup that synchronizes 360° video playback with multimodal biosensing (EEG for neural activity, EDA for arousal), enabling trial-level analysis of how specific visual features affect cognitive and physiological responses.
Cognitive & Affective Measurement Design: Integrated in-headset questionnaires and post-exposure surveys to collect comfort, stress, and perceived restorativeness ratings aligned with each stimulus, allowing us to link subjective appraisals with neural and physiological markers.
Protocol Standardization & Lab Deployment(Ongoing): Finalizing a reproducible experimental protocol and stimuli set and preparing VR lab sessions with a target sample of n ≈ 80 participants, including IRB-approved procedures for data collection, preprocessing, and cross-site analysis.

Using the resulting models to identify which combinations of greenery, enclosure, and visual complexity most strongly predict stress reduction, with the goal of informing design guidelines for health-supportive streets and public spaces.

We also tried to use Gaussian Splatting to generate hifi reconstructions of the scenes, it turned out not that well🥲.