Research On Non-contact Heart Rate Measurement Based On Face Image

Heart rate(HR)is an important physiological parameters for clinical diagnosis and treatment,which indicates cardiovascular function and human health.Traditional method of heart rate detection impedes the development of telemedicine and daily physiological parameters detection because of their inconvenient operation、expensive equipment and easily makes patients suffering from psychological burden.Compared with contact heart rate detection method,non-contact detection method based on Photo-Plethysmography Imaging(PPGI)has the advantages of noninvasive、convenient and portable.Extracting HR from face images is one of the ways to implement non-contact detection,however,HR signal acquired by capturing face video through ordinary camera is so weak that is difficult to be separated from low frequency noise such as breathing,skeletal muscle contraction and the body slightly trembling.Moreover,HR detection is easily effected by patients’ movement.Therefore,how to enhance HR signal,suppress motion artifacts and separate HR from noise is the main purpose of this study.The main work of this paper includes the following aspects:1.Analyze the waveform characteristics of PPGI signal and factors that influence the accuracy of non-contact HR detection,then proposed that using wavelet denoising and reconstruction algorithm based on Mallat algorithm to separate the HR signal from PPGI signal according to the feature of noise.2.In order to solve the problem that HR signal is too weak to be identified,this paper study the method that using Eulerian Video Magnification(EVM)to enhance color variety in HSV space,amplify the amplitude range of HR signal.Meanwhile combining with Lagrange perspective for face tracking technology to acquire Region of Interest(ROI)to reduce the motion artifacts.Experiments result showed that color enhancement combined with Lagrange perspective can suppress motion artifacts effectively and enhance HR signal even in motion state.3.Improve the face-based non-contact HR measurement process,and for the instability of HR signal,use time-frequency analysis method for periodic analysis,real-time HR value can be calculated through this method.Then this paper use Bland-Altman method to evaluate the consistency between the improved HR measurement method and traditional measurement method,and proved that the improved method is effective.4.Carried out experiments to investigate different factor’ effects on the measurement accuracy,such as video duration,different HR band,ROI,illumination,and confirmed that this detection method can applies to different scenes accurately and efficiently.