Research Of Video-Based Heart Rate Monitoring Using Joint Blind Source Separation

Heart rate(HR) is an important physiological parameter to evaluate human's physiological status.It contains abundant information to evaluate health status and disease risk.Accurate and reliable HR monitoring can prevent related diseases as early as possible,which is of great research significance.Traditional heart rate monitoring methods have many limitations due to the need for direct skin contact with sensors.Recently,remote photoplethysmography(rPPG) technology has become a research hotspot.It can continuously extract human heart rate useing ordinary cameras.Due to advantages such as low cost,simple operation,and wide application scenarios,rPPG has attracted extensive attention from researchers.This paper proposes a new rPPG method to extract human heart rate.The main work includes:Firstly,determine the facial region of interest from the captured video frames and divide it into several subregions.Two input signals are defined from each subregion,i.e.,Green channel signal or CHROM signal.Here the CHROM signal refers to a signal processed from RGB channels by a chrominance algorithm.Secondly,the empirical mode decomposition method is taken to decompose the input signals of all subregions into a number of IMFs(intrinsic mode function).The IMFs with dominate frequency within(0.7,3)Hz are retained for further process.Thirdly,two joint blind source separation methods including(independent vector analysis,IVA) and(multiset canonical correlation analysis,MCCA),are used to extract heart rate pulses with high correlation among all sub-regions using the selected IMFs.Then the energies of all candidate HR pulses are calculated,while the one with the largest proportion of the dominate and second harmonic frequencies is selected as the final HR pulse signal.Finally,the heart rate is extracted from the pulse signal by a spectrum analysis.The proposed method has been verified in a public dataset UBFC-RPPG(stands for univ. Bourgogne franche-comte Remote PhotoPlethysmoGraphy).The experimental results proves the accuracy and robustness of our method.