Remote Physiological Measurement

Remote physiological measurement (RPM) uses cameras and computer vision to non-invasively extract vital signs like heart rate from facial videos. Current research heavily focuses on improving the robustness and generalizability of RPM models across diverse environments and populations, employing techniques like test-time adaptation, domain adaptation, and state-space models (e.g., Mamba) to address challenges posed by noise, motion artifacts, and variations in lighting and skin tone. These advancements are significant for expanding the accessibility and reliability of remote health monitoring, particularly in telehealth and affective computing applications. Furthermore, research is actively exploring self-supervised and federated learning approaches to reduce reliance on large labeled datasets and enhance data privacy.