Reliable estimation of physiological indicators such as heart rate (HR) and respiratory rate (RR) from camera-based systems is important for enabling non-invasive contactless monitoring in healthcare. This work evaluates recent remote photoplethysmography (rPPG) and motion-based methods for physiological signal measurement using the OMuSense-23 database. The dataset comprises recordings from 50 participants and includes synchronized RGB-D video and ground truth photoplethysmography (GT PPG) and respiratory signals collected with medical-grade sensors. The analysis assesses traditional signal processing and deep learning-based methods in controlled breathing patterns (normal respiration, guided respiration, reading, and apnea). HR estimation is evaluated through multiple rPPG methods, while RR estimation includes two approaches: the first based on body motion tracking and the second based on derived rPPG signals. The study highlights the challenges encountered by signal processing methods under certain breathing patterns and identifies factors that affect deep learning performance, such as motion artifacts and respiratory-induced variations. These findings provide practical insights for improving non-contact physiological monitoring and enhancing real-world applications in remote healthcare.