Baby Sleep State Monitoring

Problem

Judging an infant’s sleep state from night-vision footage is hard. Parents often have to replay monitor clips, mark waking, restlessness, drowsiness, and sleep segments manually, then piece together routines. When videos are local or remote, maintaining reports and comparing historical states becomes tedious.

How It Works

The skill takes infant room monitoring video as input and runs scripts.infant_sleep_monitoring_analysis to classify deep sleep, light sleep, awakening, and restlessness. Inputs can be local video paths or network URLs, with formats such as mp4, avi, and mov.
- Sleep state classification and segment identification
- Sleep duration statistics
- Wake count statistics
- Structured sleep reports
- Routine analysis

The key workflow is: prepare monitor footage covering the crib, run the analysis script, review the structured report, identification results, and recommendation links; for historical comparison, query the cloud report inventory with --list. Historical reports must be fetched from the cloud API rather than local notes or long-term memory. --detail controls output granularity, and --output writes results to a file. User identity is handled internally, so users are not asked to provide identity parameters.

Boundaries and Notes

This is for parenting reference only and cannot replace pediatric or professional medical advice. Footage should clearly cover the whole crib, avoid clutter, and show the baby’s outline under night vision or infrared. Inputs must meet format and size limits. Network URLs are handled by the API service, so local download is not required.