Remote Photoplethysmography Using Nonlinear Mode Decomposition

Abstract

Remote Photoplethysmography (rPPG) is a contactless noninvasive method for measuring physiological signals such as the heart rate (HR) using the light reflected from the facial tissue. Signal decomposition approaches are used to extract the heart rate signal from the subtle changes in the skin color. In this paper, we show that a recently proposed signal decomposition method, namely nonlinear mode decomposition (NMD), is quite successful in estimating the heart rate signal from face videos in the presence of subject motion. Experimental results on the PureDL dataset show that NMD based HR estimation gives better results as compared to well-known methods in the literature, which use Independent Component Analysis (ICA) for signal decomposition.