People wellness monitoring

A method to assist a subject, comprising:

• capturing streaming data from one or more sensors;
• applying one or more deep learning models to identify one or more activities of the person;
• applying one or more directed acyclic graphs (DAGs) to the streaming data and the one or more activities of the person to identify a pattern from the streaming data for alert or notification of one or more predefined conditions affecting wellness;
• retraining the one or more deep learning models based on the subject's usage patterns;
• automatically adjusting one or more notification thresholds based on the learned patterns;
• and generating one or more notifications or alerts to a caregiver or a family member to assist the person based on the learned usage patterns;
• triggering exit notifications, including the steps of requiring motion in a zone prior to door open, checking for motion 30 seconds after door close, and sending a notification if no motion is detected, and checking for motion 120 seconds after door open and sending a follow-up notification if said motion is detected.