Ambient Intelligence Ecosystem for Elderly Pattern Detection and Care Using Social Robots

This article presents an ambient intelligence system to provide support and security to residents living alone, particularly in supervised domiciles. It confronts the challenges arising from the burgeoning elderly population in Europe, many of whom opt for independent living. The system combines non-intrusive sensors, a social robot, and algorithms geared towards activity monitoring and anomaly detection. Central to its functionality is the monitoring of activities, employing a hierarchical Hidden Markov Model (HMM)-based Human Activity Recognition (HAR) system. This system adeptly discerns the Activities of Daily Living (ADLs) performed by users, enabling swift interventions or alerts in instances of emergencies. Furthermore, anomalies such as prolonged periods of inactivity or nocturnal wanderings are promptly identified, triggering requisite responses to ensure user safety and well-being. A key component is the cognitive system, which establishes rules for actions predicated on data gleaned from the HAR module and anomaly detection. This enables interactions with the social robot and interventions within the environment. Validation and performance evaluation in real-world scenarios underscore the system’s efficacy, with high precision, recall, and F1Score metrics reported for the HAR module.