From pixels to a pulse signal
The camera records a rapid sequence of frames. Software identifies suitable regions of exposed skin and tracks subtle changes in colour over time. Because blood absorbs and reflects light differently as its volume changes, a repeating pulse-related pattern may be present in the colour channels.
The useful signal is small, so software must separate it from larger changes caused by head movement, facial expression, shadows, screen flicker and camera processing.
Why quality checks matter
A responsible system assesses whether the video contains enough reliable information. It can guide the user to improve lighting, hold still, remove a face covering or reposition the camera. If the signal remains inadequate, the safer outcome is no result.
Quality is not the same as clinical accuracy: it means the input was usable. Validation must separately test outputs against an appropriate reference measurement across relevant people and environments.
From pulse to different markers
Heart rate can be estimated from the frequency of the pulse signal. Beat-to-beat timing may contribute to heart-rate-variability features when sampling and signal quality support it. Breathing can influence pulse amplitude and timing, which may support a breathing-rate estimate.
Blood pressure and risk indicators require additional modelling and should be treated as distinct claims with their own evidence, limits and regulatory assessment.