Illumination-Robust Camera-Based Heart-Rate Estimation for Physiological Sensing in Robots
Researchers have developed a new transformer-based framework to improve heart-rate estimation from RGB cameras in robots, even under varying illumination conditions. This system integrates 3D face alignment, illumination augmentation, and a novel temporal standardization module. By combining waveform and spectral loss functions, the framework significantly reduces mean absolute error in heart-rate estimation, achieving a correlation of 0.982 on a new dataset designed to test illumination robustness. AI
IMPACT This research could enable more robust human-robot interaction by allowing robots to accurately sense human physiological states in diverse environments.