Research On Heart Rate Estimation Algorithm Based On Face Video

Heart rate is an important physiological index of human body,which has important guiding significance for the daily healthy life and the prevention and control of cardiovascular disease.In the process of heart rate measurement,the traditional heart rate measurement technology is mainly contact measurement and needs to be in direct contact with the subject’s skin,which is easy to bring physical and psychological discomfort to the subject.With the continuous progress of people’s health awareness,the non-contact heart rate estimation method based on face video has become a research hotspot in recent years because of it has the characteristics of comfort,convenience,flexibility,efficiency and simple operation.The method of heart rate estimation based on face video is easily affected by the interference of face tilt,motion artifact and illumination change.This thesis mainly studies the heart rate estimation algorithm based on face video to deal with the above interference.The work of this thesis is as follows:(1)Extraction of original heart rate signal based on face video.Two commonly used face detection algorithms are evaluated in terms of accuracy and speed.In this thesis,SSD face detection method accelerated by Openvino is used to improve the overall performance of heart rate estimation algorithm.Aiming at the problem of inaccurate ROI location caused by face tilt,this thesis uses LBF model to correct the face tilt,which effectively improves the accuracy of ROI selection.The skin segmentation algorithm is used to eliminate the interference pixels in ROI and form the original heart rate signal.(2)Signal denoising and heart rate calculation.Aiming at the illumination interference and motion interference in face video,NLMS adaptive filtering algorithm and PCA algorithm are proposed to reduce illumination noise and motion noise in heart rate signal respectively,which effectively improves the signal-to-noise ratio of the heart rate signal.The detrended algorithm and Butterworth band-pass filter are used to further remove the trend noise and the noise outside the heart rate frequency,and the next step is to estimate the value of heart rate in frequency domain.(3)Experimental analysis of the accuracy and applicability of the proposed heart rate estimation scheme.Comparative experiments are designed from two aspects of face tilt interference and ambient light interference.The experimental results are compared with those of different algorithms in terms of average absolute error,root mean square error,average accuracy,correlation and consistency.Experimental results show that the proposed scheme has high accuracy and effectiveness in various application scenarios.