Abstract :
[en] Vital signs are a group of biological indicators that show the status of the body’s life-sustaining functions. They provide an objective measurement of the essential physiological functions of a living organism, and their assessment is the critical first step for any clinical evaluation. Monitoring vital sign information provides valuable insight into the patient's condition, including how they are responding to medical treatment and, more importantly, whether the patient is deteriorating. However, conventional contact-based devices are inappropriate for long-term continuous monitoring. Besides mobility restrictions and stress, they can cause discomfort, and epidermal damage, and even lead to pressure necrosis. On the other hand, the contactless monitoring of vital signs using radar devices has several advantages. Radar signals can penetrate through different materials and are not affected by skin pigmentation or external light conditions. Additionally, these devices preserve privacy, can be low-cost, and transmit no more power than a mobile phone. Despite recent advances, accurate contactless vital sign monitoring is still challenging in practical scenarios. The challenge stems from the fact that when we breathe, or when the heart beats, the tiny induced motion of the chest wall surface can be smaller than one millimeter. This means that the vital sign information can be easily lost in the background noise, or even masked by additional body movements from the monitored subject. This thesis aims to propose innovative signal processing solutions to enable the contactless monitoring of vital signs in practical scenarios. Its main contributions are threefold: a new algorithm for recovering the chest wall movements from radar signals; a novel random body movement and interference mitigation technique; and a simple, yet robust and accurate, adaptive estimation framework. These contributions were tested under different operational conditions and scenarios, spanning ideal simulation settings, real data collected while imitating common working conditions in an office environment, and a complete validation with premature babies in a critical care environment. The proposed algorithms were able to precisely recover the chest wall motion, effectively reducing the interfering effects of random body movements, and allowing clear identification of different breathing patterns. This capability is the first step toward frequency estimation and early non-invasive diagnosis of cardiorespiratory problems. In addition, most of the time, the adaptive estimation framework provided breathing and heart rate estimates within the predefined error intervals, being capable of tracking the reference values in different scenarios. Our findings shed light on the strengths and limitations of this technology and lay the foundation for future studies toward a complete contactless solution for vital signs monitoring.
