Research On Non-contact Heart Rate Detection Method Based On Face Video

Heart rate(HR)is one of the important vital signs for diagnosing several diseases such as heart disease and determining the health status,and most cardiovascular diseases are preventable and monitorable.Therefore,HR measurement plays a very important role in daily medical care.Nowadays,many contact measuring instruments are being widely used due to their high accuracies,but they must be in close contact with human skin,which makes the subject uncomfortable,and is not suitable for long-term measurement.Furthermore,they cannot be used for special people(for example,those who are suffering from infectious diseases or allergic skin diseases,or those who are suffering from burns,etc.).Imaging photoplethysmography(iPPG)is a non-invasive,non-contact HR measurement technology,which uses the ambient light as a source of illumination and uses a camera without contacting the sensor on the skin to record the color change of the skin exposed to the human body to measure the HR.This technology is very suitable for daily health monitoring due to its advantages such as no contact with the human body,simple operation and low cost.However,this technology is susceptible to the interference of the ambient illumination variations and motion artifacts.In addition,among the previous iPPG-based HR measurement researches,most of them have only focused on the interference of illumination variations or motion artifacts,while few studies have considered both interference of illumination variations and motion artifacts.Based on the systematic analysis of the theoretical basis of iPPG and the existing iPPG-based HR estimation research results,this paper proposed a HR estimation method that is robust to the interference of illumination variations and motion artifacts.The main research contents of this paper are as follows:(1)On the basis of summarizing the current development status of HR estimation methods based on the iPPG technology at home and abroad,the challenges faced in this research field,the quantitative and qualitative evaluation indicators of the performance of HR estimation method,and the characteristics of the commonly used public databases were given.(2)The method of combining different color channels was studied.This article first summarized the commonly used color space and color channel combination methods in the field of iPPG-based HR estimation.Then,it selected 13 different color channels and their combination methods,and compared them using the public databases UBFC-RPPG and PURE.After that,the combination method that contains the most blood volume pulse(BVP)signals has been confirmed.(3)On the basis of determining the channel combination method that contained the most BVP signals,the POS-based method was improved,and a HR estimation method that used the signal of the skin region to directly suppress the interference of illumination variations without using the background signal was proposed.In order to verify the effectiveness of this method,this paper collected the data of 13 subjects under 21 different illumination conditions,and by using the collected data and the public database UBFC-RPPG,this method was compared with other methods that have the abilities to suppress the interference of illumination variations.The results showed that the performance of this method was better than other methods compared,while the mean errors,mean absolute errors,root mean squared errors,accuracies,Pearson correlation coefficients and 95 % limits of agreement for the data collected under different lighting conditions and the public database UBFC-RPPG were 0.44 bpm and 0.58 bpm,1.81 bpm and 1.75 bpm,4.41 bpm and 2.40 bpm,93.0 % and 95.8%,0.91 and 0.99,-8.21 bpm～9.09 bpm and-4.13 bpm～5.29 bpm,respectively.(4)On the basis of the research on suppressing the interference of illumination variations,through the analysis of the characteristics of each model-based method,a HR estimation method was proposed that was robust to the interference of illumination variations and motion artifacts.And then,this paper used the data of 13 subjects collected under the three different conditions and the public database PURE to verify the effectiveness of the proposed method.The results showed that the method proposed in this paper had better robustness to the interference of illumination variations and motion artifacts than the compared method,while the mean errors,mean absolute errors,root mean squared errors,accuracies,Pearson correlation coefficients and 95 % limits of agreement for the data collected under the different illumination conditions and the public database PURE were 0.595 bpm and 0.034 bpm,1.135 bpm and 0.448 bpm,1.624 bpm and 1.025 bpm,100 % and 100 %,0.991 and 0.9991,-2.59 bpm～3.78 bpm and-1.97 bpm～2.04 bpm,respectively.