A video-based non-contact heart rate measurement method is proposed.
The method jointly analyzes temporal colour channel signals of facial sub-regions.
A max-margin multi-label classifier is used to predict optimal dataset correlation.
A large public database is used to verify the performance of the proposed method.