Understanding natural human behavior in daily living scenarios is crucial for assessing and monitoring neurological and age-related functional declines. However, conventional observational and behavioral studies often lack ecological validity and control in addressing the complex real-world challenges of analyzing mobility, cognition, and behavior. In this work, we present an AI-enhanced virtual reality system for behavioral analysis of elderly people. Our system simulates high-fidelity daily living environments incorporating dynamic weather and traffic conditions, and records 3D human behavior with motion capture technology. Multimodal analysis revealed the influence of dynamic and adjustable VR interactions on behavioral changes in individuals at risk of neurological declines, underscoring its potential for assessing and detecting motor and behavioral dysfunctions in aging and neurological populations. Our findings also highlight the capability of our VR-based system as a platform for studying natural human behavior and cognitive-motor integration, and advancing AI-driven behavioral research and health assessments.