Research Of Facial Video-based Pulse Measuring

The heart rate, as the body’s most important, and one of the most basic physiologicalparameters, has important implications on the health indications. The traditional way ofmeasurement requires the measure contact with the doctor or the instrument, to some extentcaused by the measurement inconvenience. This paper discusses a measurement which bases onphotoplethysmography principle, using computer vision technology and independent componentanalysis to deal with a clip of human face video. This is a non contact measurement of the pulsemethod, and the paper is divided into three parts:Firstly, obtain the video data. using computer vision technology separates the clip of humanface video, in the RGB color space model, into three color channel, and take the average of itsspace as the original data.Secondly, the independent component analysis processing. Based on the non Gaussian theindependence of judgment in information theory and the characteristics of the ICA algorithm,The paper process the original data with the analysis of FastICA algorithm. First at all, it is thedata preprocessing, including zero mean, bleaching, and then select the objective function, theICA algorithm deal with the pretreatment data, separation of independent conmponent, and thenmake filting the data with an arithmetic meam filter,as time as handle detrending on the data,and last select one independent as the heart rate.Thirdly, consistency with the traditional method. Since this experiment is a newmeasurement of the pulse method, last the paper prove that the measurement has someconsistency with the traditional way of measurement using the Bland Altman method.The experimental results show that proposed method is feasible and effective, and has acertain practical value and application prospect in daily life and telemedicine.