Fall Risk and Gait Analysis in Community-Dwelling Older Adults using World-Spaced 3D Human Mesh Recovery

A team of researchers, including Chitra Banarjee, Patrick Kwon, and colleagues, has developed a novel AI pipeline that could revolutionize how fall risk is assessed in older adults. Presented at the Computer Vision for Biomechanics Workshop at CVPR 2026, the system leverages a 3D Human Mesh Recovery (HMR) model—a computer vision technique that reconstructs a detailed 3D model of a person's body pose and shape from a 2D video. This model analyzes recordings of individuals performing the standard Timed Up and Go (TUG) test, a common clinical assessment where a person stands from a chair, walks a short distance, turns, walks back, and sits down.

From these simple videos, the AI extracts precise spatiotemporal gait parameters that are difficult for clinicians to measure manually, including step time, step length variability, and sit-to-stand transition duration. Crucially, the study validated the AI's outputs against gold-standard sensor data. They found that step times calculated from video were significantly correlated with measurements from instrumented insoles containing Inertial Measurement Units (IMUs). Furthermore, their statistical models confirmed that the AI-derived metrics—specifically shorter, more variable step lengths and longer sit-to-stand times—were strong predictors of an individual's self-reported fear of falling and perceived fall risk.

This research demonstrates a major step toward 'ecologically valid' assessment. Instead of requiring a lab visit with specialized, expensive equipment like motion capture suits or pressure-sensitive walkways, this method enables accurate gait analysis in real-world community settings like senior centers using only a smartphone or tablet camera. The pipeline automates the extraction of clinically meaningful biomarkers, moving beyond the simplistic stopwatch-measured gait speed that is the current standard of care. This has the potential to make frequent, low-cost, and widespread fall risk screening a practical reality, allowing for earlier interventions.