DefinePK

Abstract

The increasing demand for continuous, real-time health monitoring has driven advancements in intelligent embedded systems that integrate physiological sensing, machine learning, and sensor fusion. This study presents the design and evaluation of a smart embedded system capable of capturing and classifying multiple physiological signals—including heart rate, SpO₂, body temperature, respiration rate, and activity level—for early detection of health anomalies. A suite of machine learning models, including Logistic Regression, Random Forest, Support Vector Machine (SVM), Convolutional Neural Network (CNN), and K-Nearest Neighbors (KNN), were trained and tested using features extracted from the fused sensor data. CNN demonstrated the highest classification accuracy (93.5%), while Logistic Regression recorded the best AUC (0.80), highlighting different strengths across models. Feature importance analysis revealed heart rate variability (HRV), SpO₂ mean, and temperature trend as the most influential predictors. Additionally, correlation analysis emphasized the synergistic relationships between physiological parameters, reinforcing the value of sensor fusion in signal interpretation. The proposed system offers a portable, efficient, and scalable solution for real-time physiological monitoring, with potential applications in remote healthcare, fitness tracking, and wearable technologies