VOR-rPPG (Variance-Optimized Remote Photoplethysmography) is a high-performance C++20 framework for robust, real-time heart rate estimation from video sequences. Remote photoplethysmography (rPPG) methods are inherently sensitive to motion artifacts and illumination variability. This framework addresses these challenges through a novel architecture that combines multi-modal signal fusion, risk-aware spectral weighting, and temporal stability control. Key Features: • Multi-Modal Signal Fusion:Integrates candidate signals derived from POS (Plane-Orthogonal-to-Skin), CHROM (Chrominance-based), and Green-channel methods using a greedy clustering strategy to maximize signal reliability. • Risk-Aware Weighting (φ):Introduces a sigmoid-based confidence scoring function that penalizes signals near spectral boundaries and those affected by harmonic interference, improving robustness in noisy conditions. • Stability State Machine:Implements a temporal trust mechanism that ensures smooth heart rate estimation and prevents abrupt fluctuations caused by transient noise. • High-Performance C++20 Architecture:Designed with modern memory management and zero-cost abstractions, enabling efficient real-time execution and suitability for embedded systems. Performance: The system supports multi-ROI (Region of Interest) processing, allowing redundant spatial sampling to improve signal-to-noise ratio (SNR). This makes it suitable for applications in telemedicine, non-contact vital sign monitoring, and human-computer interaction. Open Science: The full source code, documentation, and benchmark examples are publicly available: GitHub Repository:https://github.com/Zyreniee/VOR-rPPG License:MIT License (software)CC-BY 4.0 (documentation/publication)
Gön Yusuf (Thu,) studied this question.