Elderly Gait Instability and Fall Risk Detection

Problem

Older adults may show gait changes before falls, but home and care settings often lack quantified screening. This skill targets straight-line walking videos from a fixed camera in hallways or living rooms, focusing on detecting short shuffling steps, slow gait speed, and increased trunk sway—not making medical diagnoses.

How It Works

• Input: mp4/avi/mov video, preferably 10–30 seconds, at least 25 FPS, covering 3–5 continuous steps; height may be supplied for pixel-to-centimeter conversion.
• Metrics: step_length_cm, gait_speed_m_s, cadence_steps_min, trunk_sway_deg, step_length_variability, and double_support_ratio.
• Patterns and risk: classifies normal / short_steps / wide_sway / slow / mixed and outputs low / medium / high fall-risk levels with key risk factors.
• Reports: produces structured reports and can query cloud history with --list, rendering reportImageUrl links in a Markdown table.

Boundaries

• Measurements are affected by camera angle, lighting, clothing, and calibration; without height, absolute centimeter values are mainly useful for trend comparison.
• Output is an auxiliary screening aid, not a substitute for neurology, rehabilitation, or clinical evaluation.
• Gait video contains sensitive health data; obtain informed consent from the person or family and protect or encrypt recordings.