Study On RPPG Based Heart Rate Measurement Algorithms For Non-Ideal Illumination

Remote photoplethysmography(rPPG)enables non-contact heart rate measurement by extracting the subtle color changes in videos of human’s bare skin under ambient illumination.In general,rPPG requires ideal illumination environments with suitable and constant brightness with stable spectral composition,so its application under practical non-ideal illumination environments will be limited.This paper studies the rPPG based heart rate measurement algorithm through videos under non-ideal illumination environments,and proposes rPPG based heart rate measurement algorithms with high robustness under different non-ideal illumination environments.The main research contents of this paper are as follows:(1)Based on the model of rPPG,the influence mechanism of multiple non-ideal illuminations on the pulsatile signal extracted by rPPG is analyzed,and the limitations of the traditional rPPG algorithm under non-ideal illumination conditions are explained.(2)Research on rPPG based heart rate measurement algorithm in low-light environments.Aiming at the characteristics of weak color change caused by pulsation in low-light environments,the Eulerian video magnification(EVM)is proposed to enhance the color change signal within the frequency range corresponding to normal heart rate and realize the detection of faces in low-light environments by the low-light image enhancement algorithm.Experiment results show the effectiveness of the proposed method for rPPG based heart rate measurement in low-light environments.(3)Research on rPPG heart rate measurement algorithm under illuminant-varying environments.Aiming at the environment of varying illuminant intensity and varying illuminant color,the corresponding algorithm is proposed to remove the influence of changing illuminant.For the varying intensity environment,the independent component analysis(ICA)is used to separate the pulsatile component induced by heart rate and the light intensity change in each color channel,then the interference component left in the pulsatile component is removed by high-pass filter to achieve the suppression of the interference of the varying illuminant intensity.For the varying illuminant color environment,firstly the RGB color space is converted to the LAB color space.Then the L channel representing luminance with more interference is removed,and the A and B color channels are analyzed by singular spectrum analysis to remove the trending term to obtain the pulsatile signal.Experiments show that the proposed methods greatly improve the accuracy of the heart rate measurement algorithm in the environment of varying illuminant intensity and varying illuminant color.(4)Multi-region heart rate measurement algorithm based on rPPG.By analyzing the characteristics of rPPG pulse signals in different regions of the human face,the distance from the RGB color channel signal difference dispersion point map to the pulsatile signal direction is proposed as the basis for region of interest(ROI)selection.Combined with the joint blind source separation,the multi-region measurement of heart rate is realized,which further improves the accuracy of the rPPG heart rate measurement algorithm in non-ideal illumination environments.