Research On Heart Rate Detection Based On Face Video

Resting heart rate is an important indicator of cardiovascular health.At present,the methods for detecting heart rate mainly include contact and non-contact detection.The contact method requires the instrument to directly touch the human skin,which causes discomfort and inconvenience.Non-contact measurement is a new type of measurement method,which has the advantages of simple operation,high comfort and continuous measurement.It is very suitable for daily health care,telemedicine and intensive care.This thesis is based on photoplethysmography,to study the method of ROI(Region Of Interest)selection based on face video heart rate detection.We propose two methods of selection: active method and adaptive method to obtain effective heart rate information from video.The main contributions of this thesis are as follows:1)It summarizes the principles of the two types of methods based on video heart rate detection technology,current research development and existing problems by reviewing literatures.On this basis,the motivation of solving problems is obtained.2)This thesis classifies the selection ROI methods.Inspired by the existing research,we propose an active ROI selection method.The method selects the centroid blocks in the cheek region as the signal processing area by locating the feature points,which is a low complexity and feasible regional selection method.3)To overcome the drawback of active selection method,we propose an adaptive region selection method,in which the characteristic signal matrix is constructed to decompose the noise.The region of interest with high SNR can be adaptively selected.The method has high accuracy and still has good robustness under the real environment.We conduct experiments on 15 subjects to evaluate the effictiveness of the proposed methods.The experimental results demonstrate both the active selection method and the adaptive selection method achieve an accuracy of over 94%.The experimental results show that the ROI selection method in this thesis has good feasibility in the video heart rate detection technology.