Non-contact Heart Rate Detection Based On Skin Segmentation And Eulerian Video Magnification

Non-contact heart rate detection based on remote photoplethysmography(r PPG)technology has a wide application prospect in telemedicine and daily physiological parameter detection due to its simple,easy-to-operate,non-invasive,and low-cost advantages.However,the non-contact heart rate detection based on r PPG technology is founded on the subtle color change of the human skin due to the heartbeat.Non-skin pixel interference,weak skin color change,and motion artifact noise will all affect the heart rate detection results.Therefore,the study of skin segmentation and color enhancement is important for accurate non-contact measurement of heart rate.In this article,a non-contact heart rate detection method is designed and implemented based on skin segmentation and Eulerian video magnification(EVM).The validity of the proposed method is verified by experiments.Firstly,the skin segmentation model is used to segment the skin area in the video.Next,for the segmented skin,EVM algorithm is used to magnify the weak skin color change signal caused by the heartbeat.Then,the source signal is separated from the magnified color change signal through independent component analysis algorithm,and the source signal that characterizes the heart beat is selected as the r PPG signal by using the correlation analysis.Finally,the heart rate is estimated by analyzing the power spectral density of the r PPG signal.The method proposed in this thesis is verified on the public-domain UBFC-RPPG dataset and self-collected dataset,and the results show that the proposed method can effectively estimate the heart rate.The main work of this thesis is as follows:1.A skin segmentation model based on Fully Convolutional Networks(FCN)is proposed,and its performance is verified experimentally,and the 90.07% Mean Intersection Over Union(MIOU)is obtained.2.On the public-domain UBFC-RPPG dataset,the performance of the non-contact heart rate detection method proposed in this thesis is studied from two aspects: skin area selection method and r PPG signal extraction algorithm.Experiments prove that the method proposed in this article has better accuracy in heart rate detection and is more robust to interference with non-skin pixels than traditional methods.It also proves that using EVM algorithm to magnify the faint color changes of the skin can improve the accuracy of heart rate detection to a certain extent.3.On the self-collected dataset,the robustness of light source change,light condition change and occlusion are studied,and the validity of the non-contact heart rate detection method in this thesis is verified by the further experiments.