Remote Photoplethysmography measurement using constrained ICA

Remote Photoplethysmography (rPPG) is a technique that consists in estimating physiological parameters such as heart rate from live or recorded video sequences taken by conventional camera or even webcams. This technique is increasingly used in many application fields thanks to its simplicity and affordability. The basic idea is that the arterial blood flow shows regularity due to the heartbeat. This regularity is manifested by very small periodic variations in the color of the skin, which can be isolated and quantified by signal and image processing methods. In this context, Independent Component Analysis (ICA) is largely used to separate the signal due to arterial flow from signals from other sources (skin, lighting, etc.). However, basic ICA is considered blind, i.e. it uses no a priori knowledge of the sources leading to issues in identification of the separated sources. We propose the constrained ICA (cICA) method where we take advantage of the knowledge about the periodicity of the blood flow signal along with the CHROM constraint. The periodicity is implemented by means of autocorrelation maximization and the CHROM constraint helps to automatically set some parameters. We tested our method with the MMSE-HR database for the measurement of rPPG where it showed better performance compared to conventional ICA and other state of the art methods in terms of accuracy and robustness.