A comprehensive survey on contactless vital sign monitoring using vision-based, radio-based, and fusion approaches

Continuous monitoring of vital signs, such as heart rate (HR), respiratory rate (RR), blood pressure (BP), and body temperature (BT), is important in both clinical and non-clinical settings. However, traditional contact-based methods are often unsuitable in contexts such as neonatal care, burn units, or during pandemics due to discomfort, hygiene concerns, and patient non-compliance. This paper presents a comprehensive survey of contactless vital sign monitoring systems based on vision and radio-frequency (RF) technologies, covering RGB, depth and thermal imaging, as well as radar-based sensing including millimeter-wave systems. We critically compare the strengths, limitations, and environmental constraints of single-modal (vision or RF) and multimodal (radio-visual fusion) approaches, and detail common fusion strategies at feature, intermediate, and decision levels. The review includes a curated taxonomy of state-of-the-art algorithms for HR, RR, BP, and BT estimation, together with an analysis of public datasets and benchmark evaluation metrics. Based on a narrative review with iterative searches across major peer-reviewed databases, covering over 150 studies from 1975 to 2025, we identify technical gaps, particularly in non-contact BP and BT estimation, multi-person monitoring, and the lack of standardized validation protocols, and we highlight open challenges related to motion robustness, demographic variability, privacy, and clinical translation. The survey also links these methods to application drivers across hospital monitoring and telehealth, home-based wellness, and adaptive human-centered systems such as XR experiences that benefit from continuous physiological and affective state estimation. This survey bridges sensing hardware, signal processing, and machine learning, and outlines research directions toward robust non-contact health monitoring systems suitable for practical deployment.