MPU-rPPG: A Comprehensive and High-Fidelity Dataset for Remote Photoplethysmography Across Diverse Conditions and...

Remote photoplethysmography (rPPG) is a non-contact singal extracting photoplethysmographic (PPG) waveforms from video by analyzing blood-volume-induced modulations in the light reflected from the skin. As rPPG moves closer to real-world deployment, comprehensive datasets are essential for the development of robust algorithms across diverse conditions and populations. Building on existing efforts, we introduce the MPU-rPPG dataset, designed to bridge current gaps and establish a new gold standard for rPPG research. This dataset captures high-fidelity physiological signals from multiple body regions (including the face and limbs) under varying light conditions and environments. It supports heart rate ranges from 50-160 bpm, accommodating both resting states and high-intensity activities, and includes a diverse set of subjects across different skin tones and demographics. The versatility of MPU-rPPG supports robustness studies relevant to practical scenarios-including clinical monitoring environments and highly dynamic public activities (e.g., endurance sports or driving)-where motion, illumination variation, and partial occlusion remain major challenges for reliable rPPG. By addressing these critical needs, the MPU-rPPG dataset provides a comprehensive foundation for the development of real-world rPPG systems and is committed to setting a new benchmark for future research.