Heart rate(HR)parameter is an important physiological indicator reflecting the health of the human body,and it has a huge application prospect in the field of medical.Accurate and effective heart rate monitoring can not only prevent the occurrence of related diseases as soon as possible,but also provide a basis for later treatment.Recently,the researchers have invented a non-contact measurement method called Image Photoplethysmography(IPPG).This method collects facial skin images through imaging equipment,analyzes the color changes on the skin surface to extract the original signal,and finally extracts effective heart rate information through a certain denoising algorithm.Compared with the traditional contact measurement method,this method has lots of advantages such as comfortable,non-invasiveness and cheap.However,the head movement of the subject and lighting conditions severely limit the wide application of this method.In order to solve these problems,this paper designs and implements a HR detection system that resists small movements and a HR detection system based on near-infrared images.The main work of this thesis is as follows:In the video-based non-contact HR detection method,the choice of different regions of interest will affect the results of the heart rate measurement.To solve this problem,based on the distribution of facial blood vessels and actual application requirements,this thesis proposes a method of extracting heart rate information from the nose area.Experiments show that compared to other areas,the nose can meet actual application requirements and effectively reduce noise.In addition,the head movement of the subject will result in inaccurate measurement results.To solve this problem,the thesis is based on the theory that stable tracking of the region of interest can effectively improve the signal-to-noise ratio of the heart rate signal.It adopts the idea of detection instead of tracking,and applies facial landmark detection algorithm to track the region of interest stably.Experiments show that the algorithm based on facial landmark detection can stably track the region of interest when the subject is moving the head,and the measurement is accurate.The accuracy rate can reach about86% at a distance of 4 meters.In dark scenes,heart rate detection methods based on color images cannot extract heart rate information.In response to this problem,the thesis adopts the theory that the near-infrared image can collect faces in dark scenes and the distribution of facial blood vessels.It proposes a method for extracting heart rate information based on the bridge of the nose and left and right cheeks.The experimental results show that the method based on the combination of multiple groups of regions of interest can significantly improve the signal-to-noise ratio of the heart rate signal.In response to this problem,this paper proposes a strategy that combines the empirical mode decomposition algorithm and the independent component analysis algorithm.Experimental results show that this method can filter out motion noise very well.The detection accuracy can reach about 96% in a dark scene at close distance. |