Investigation on different color spaces for non-contact heart rate monitoring using facial videos

Current image-based photoplethysmography (iPPG) methods depend on standard RGB face video recordings to monitor non-invasive heart rate (HR) data. This study focuses on HR monitoring, notably the effectiveness of individual single-color channels (SCC) and the combination of multiple-color channels (MCCs) in HR extraction. The approach entails decomposing different color channels into intrinsic mode functions (IMFs) via the ensemble empirical mode decomposition method. Selected IMFs with heart-related information are reconstructed, and principal component analysis is used to identify the common blood volume pulse (BVP) signal. The first primary component, with the highest cyclic changes indicating cardiac information, is chosen as the target iPPG signal. The experiment is conducted on three datasets. Results indicate that the combination of MCCs achieves better agreement between estimated and ground truth PPG values compared to SCCs, highlighting its efficacy in HR estimation.