AI-driven IoT framework for smart health monitoring using embedded sensors and edge analytics

Given the rise of chronic ailments and healthcare moving outside the purview of hospitals, there remains an unmet requirement for monitoring that is reliable, preserves privacy, and has low-latency in real time. This paper focuses on an AI-enabled IoT architecture for health monitoring that conducts edge analytics. Multimodal signals from an Empatica E4 wearable device (blood-volume pulse, ECG, skin temperature, and tri-axial acceleration) are continuously transmitted over BLE to a Raspberry Pi gateway. Upon synchronization to 4 Hz, denoising, and calculation of blood volume pulse, 10-s windows are then processed through a lightweight pipeline that includes compact statistical, temporal, spectral, and HRV features and a classifier based on a 1D-CNN plus LSTM architecture in TensorFlow Lite format post-training quantization. Cloud services store synopses of the data, visualize growth patterns, and facilitate model updates while keeping the raw data locally. On the PhysioNet Wearable Device Dataset, subject-wise 5-fold fashion yields 96.4% accuracy, 95.7% precision, 96.1% recall, and 96.0% F1 (macro-AUC ≈ 0.955). Triggered edge inference is achieved in less than 100 milliseconds per window at about 2.8 Watts. It is shown that edge artificial intelligence increases real-time monitoring capabilities and can be expanded to other biosignals and clinical areas.