Development of an AI-Driven Multi-Sensor Health Monitoring System for Real-Time Physiological Analysis and Insight Generation

Abstract

The existing healthcare systems are based on a limited number of physiological parameters, such as the body temperature, the pulse rate and the body weight, as the initial screening parameters. Even though these indicators provide an approximate evaluation, they do not reflect the more in-depth changes of physiological variations that are associated with dangers of underlying health problems, leading to delayed diagnosis and reduced effectiveness of preventive measure The present paper describes an AI-based multi-sensor health monitoring system that will create advanced physiological information through non-invasive measurements. The system combines photoplethysmography (PPG) based on the infrared, temperature, and voltage-based signal detection to obtain the most important health indicators such as heart rate, oxygen saturation, perfusion index, level of hydration, stress index, and blood pressure dynamics. Filtering, normalization and feature extraction are signal processing methods used to guarantee reliable interpretation of the data. It uses a hybrid multi-model AI structure, so that a multi-domain classification and baseline estimation is done using a Random Forest model, and high-accuracy blood pressure prediction is done using a LightGBM model. These models are linked together in a cascaded chain of inferences to improve the predictive accuracy. Also, a Large Language Model (LLM) is introduced to transform the numerical results into structured topics that can be analyzed by humans based on clinical interpretation. The suggested system has a high predictive accuracy and has a low computational overhead, allowing real-time deployment of the edge and offering a scalable and cost-efficient solution to intelligent and preventive healthcare monitoring.