Smart City Healthcare: Non-Contact Human Respiratory Monitoring With WiFi-CSI

Respiratory monitoring is vital for health applications, prompting the development of intelligent solutions for human well-being. This study focuses on WiFi-based methods for respiratory monitoring, especially relevant to consumer electronics applications. Despite their convenience and safety, existing WiFi-based approaches face challenges in accurately estimating respiratory frequency and classifying respiratory patterns. To address these issues, we propose a respiratory monitoring system consisting of commercial wireless router and PC. By leveraging Fresnel zone theory to understand the relationship between human respiratory transmission paths which provides insights into signal propagation characteristics, we can enhance the accuracy of respiratory frequency estimation and improve classification performance. We introduce a series of algorithms including an improved wavelet algorithm to achieve these goals. The maximum estimation error is 3.9 percent in comparison experiments with a baseline contact measurement device. Additionally, we present a method for constructing breathing pattern recognition features and propose a classification network tailored to breathing characteristics, enabling the recognition of normal, paused, and deep breathing patterns. Experiments are conducted in three scenarios: corridor, conference room, and laboratory, and the results show that the average classification accuracy of these three scenarios is 96.67%.