CareSight: A Convergent AI, Augmented Reality, and IoT Framework for Predictive Elderly Healthcare Monitoring

Abstract

The rapid growth of the global elderly population has intensified the need for affordable, convergent technology that supports safety, autonomy, and continuous health supervision. This paper presents CareSight, an AI-powered augmented-reality assistive and IoT healthcare monitoring system that unifies real-time hazard detection (YOLOv8), AR navigation (ARCore/ARKit), OCR text reading, voice interaction, and a wearable IoT wristband for physiological monitoring within a single smartphone-centric platform. Wristband sensor data is streamed via an ESP8266 module over MQTT and analyzed by a two-layer LSTM network for predictive anomaly detection and automated emergency alerting. The system is organized as a four-layer perception–processing–application–cloud architecture so that hazard detection, navigation, voice/OCR interaction, and health prediction share a single alerting pipeline rather than operating as separate applications. Evaluated on 5,000 annotated hazard images and 15 days of physiological data from 30 elderly volunteers, CareSight attains 94.2% object-detection accuracy at 30 FPS and 95.4% health-anomaly prediction accuracy, cutting emergency response time from 45 seconds to 8 seconds. Beyond the technical validation, the study quantifies deployment cost, scalability, and business viability, showing that the proposed phone-plus-wristband architecture costs a fraction of proprietary AR and medical-grade alternatives while remaining deployable in urban and semi-urban care settings. Taken together, the results indicate that convergence of perception, prediction, and low-cost hardware — evaluated jointly with cost and adoption criteria — is a practical, near-term path to scalable elderly-care technology rather than a purely aspirational one.