Remote photoplethysmography · rPPG
Blood volume changes skin colour.
With each heartbeat, more blood reaches the skin and absorbs slightly more light, especially green light. Across many video frames that faint change forms a pulse wave. Researchers showed in 2008 that ordinary ambient light and a consumer camera are enough. CUORE uses a neural rPPG model, FacePhys, through the open-source open-rppg library, and reports its signal quality alongside every reading.
Verkruysse, Svaasand & Nelson, “Remote plethysmographic imaging using ambient light”, Optics Express, 2008.
Optical flow
Motion between frames reveals breathing.
Optical flow estimates how each small patch of an image moves from one frame to the next. CUORE measures the vertical motion of the area below your chin with Farnebäck's dense method, then finds the dominant slow rhythm in the last 30 seconds. If that rhythm isn't clearly above the noise, no number is shown.
Farnebäck, “Two-frame motion estimation based on polynomial expansion”, SCIA, 2003.
Facial landmarks
Eyelid geometry counts blinks.
A face-landmark model places 478 points on the face in every frame. The distance between eyelid points relative to eye width gives an openness ratio, which drops sharply during a blink. CUORE measures your own open level and counts a blink when openness falls below 75 % of it for a few frames.
Soukupová & Čech, “Real-Time Eye Blink Detection using Facial Landmarks”, CVWW, 2016.