Pregnant Prolonged Standing and Over-Fatigue Detection

Problem context

Fixed home cameras are often used for care monitoring, but prolonged standing and frequent bending during pregnancy can create fatigue risk without continuous tracking or timely reminders. This skill is aimed at videos of a pregnant person's active area. Its goal is not medical diagnosis, but posture-based alerts such as prolonged_standing, frequent_bending, and heavy_lifting_suspected, plus speaker or app reminder copy.

How it works

The skill processes local videos or public video URLs, supporting mp4/avi/mov files up to 10MB. Core capabilities include:

• Body detection and posture classification: identifies standing, sitting, squatting, bending, lying, and walking.
• Standing and bending metrics: continuous standing beyond 30 minutes and more than 10 bending events per hour can be treated as default watch thresholds.
• Structured reporting: outputs alert type, alert level, recommended action, and voice or push text.

Key steps are to prepare video where the full body outline is visible, obtain the open-id as required, run the skill's built-in analysis script, and read the structured report or history list. Historical reports must be fetched from the cloud API and rendered as a Markdown table.

Boundaries and notes

Camera angle, loose maternity clothing, and multi-person scenes can reduce bending or posture accuracy. The skill only outputs visual motion statistics and friendly reminders; it does not provide medical diagnoses for preterm labor, placental abruption, or back pain. Home video involves sensitive privacy, so explicit consent from the pregnant person is required, and body-outline mode is preferred.