| Heart rate is of vital importance to human life feature and heart rate indicators are directly related to the physical health of the human body.In order to solve the problems of the inconvenience of the equipment caused by the contact with the subject and the limitation of the application caused by the test conditions of the traditional heart rate detection method,this paper studied a non-contact heart rate detecting algorithm using image photo-plethysmography technology(IPPG).Compared with the traditional heart rate detection algorithm,this measurement system has a wider application scenario and a more convenient test environment,which greatly reduces the manpower and equipment cost caused by the system.Meanwhile,its completely non-invasive feature will also significantly reduce the psychological pressure of tested subjects.Compared with the infrared detection system that also does not require direct contact to measure heart rate,this system reduces the cost of acquisition and analysis by immense scale.As a measurement system that takes both testing convenience and application cost into consideration,the noncontact heart rate measurement system that uses ordinary optical cameras to recognize faces signals needs to solve many problems,including but not limited to changes in ambient light interference,face movement and occlusion,low-frequency interference of physiological activities on the surface of human skin,etc.The main goal of this paper is to improve the current non-contact heart rate detecting method based on Image Photoplethysmography technology with the addition of external ambient light compensation to achieve a better recognition efficiency and higher testing accuracy.The improved method of this article includes the following aspects:1.Analyze the interference factors that affect the accuracy and efficiency of the heart rate detection from the optical principle of the IPPG algorithm,and conduct a step-by-step research on the possible effects of the image signal collecting,processing and analysis steps,compare it with the currently proposed non-contact heart rate detecting methods step by step,and explore the possibility of improvement.2.Aiming at the problems of missed and wrong detection in traditional face recognition methods within complex scenes,this paper proposed to replace the Adaboost face detecting method based on Haar features with Multi-task Cascaded Convolutional Networks(MTCNN)to perform the face recognition and tracking.3.In view of condition that the region of interest(ROI)for heart rate measurement is produced only by cropping the face area in current non-contact heart rate measurement method,this paper improved the relative consistency of the detection area and reduced the incoherence of pixels in the ROI.Those procedures are designed to erase the possible interference during face recognition.4.Taking the influence that external ambient light might have on the imaging quality and photoelectric volume changes into account,the normalized least mean square(Normalized Least Mean Square,NLMS)adaptive filtering algorithm is used to compensate for the external interference caused by the fluctuation of the light intensity,and then the target image signal reconstructs the heart rate information.5.The difference in recognition accuracy between the Adaboost face classifier based on Haar features and MTCNN is compared and analyzed,and the influences of different ROI that system is using to detect heart rate are discussed in term of the quality of reconstruction of heart rate.This paper also discussed the degree of influence of the processing priority of the light compensation module on the monitoring of the heart rate.6.It is proved that the improved non-contact face heart rate detection system has a certain improvement in recognition accuracy compared with the currently proposed non-contact measurement method by experiments.A real-time detection and heart rate display system is built based on those methods. |
PDF Full Text Request