AI-Driven Safety and Health Monitoring Systems: A Review of Adaptive Learning, Multimodal Sensor Fusion, and Predictive Intelligence

Abstract Research on safety and health monitoring includes wearable fall detection, human activity recognition, computer vision, connected sensors, and chronic-disease monitoring. These areas share an interest in identifying events that may require assistance, but results from an individual dataset or task do not demonstrate that a complete emergency-response system will work reliably. This paper describes LENORA, a proposed AI-assisted monitoring system covering child safety, women’s safety, elderly monitoring, and adult disease-care monitoring. Its intended workflow links category-specific inputs with event analysis, risk assessment, dashboard support, safe-word verification, and location-aware communication. The literature summaries supplied for this review highlight practical concerns such as class imbalance, privacy, energy use, and deployment constraints. Project notes estimate that the software foundation is approximately 80% complete; this is a progress estimate, not a model-performance result. The available material does not verify that all sensors, AI models, GPS, GSM/SIM800L communication, and emergency-response steps have been integrated and tested. For that reason, this paper reports no LENORA accuracy, precision, recall, F1-score, latency, or deployment results. Instead, it records the current evidence boundaries and outlines tests for classification quality, false alarms, response time, communication reliability, and end-to-end behavior. Keywords wearable sensing, fall detection, Internet of Things, edge AI, risk assessment, emergency alerting, health monitoring