Research On Non-contact Heart Rate Detection System Based On Video Analysis

Heart rate,as an important physiological parameter of human,plays an irreplaceable role in human health monitoring.This paper designs a non-contact measurement device that uses video image processing to detect heart rate.Combining embedded image processing and heart rate detection can achieve real-time acquisition of video images and amplify the weak human vital signal,and then obtain the heart rate value through signal analysis.The noncontact heart rate measurement method has the advantages of non-invasiveness,convenience,and long-term continuous measurement,which lays the foundation for the development of remote monitoring in smart medicine.The paper first studies and simulates the model of small signal amplification in video images.A mathematical model of Euler video amplification is established,and the selection of the main parameter amplification factor in this model is studied.Gaussian pyramid decomposition technology was used to decompose the video sequence into different frequency bands.Using the video image sensory contrast method and the method of calculating MSE and PSNR,it is concluded that the four-layer image pyramid is the optimal number of layers for amplifying the heart rate signal.The heart rate signal is extracted through a time-domain bandpass filter,and after being amplified,it is synthesized by Laplacian pyramid and superimposed with the original video sequence to obtain an amplified output video.Aiming at the problem that the brightness of the light affects the measurement of heart rate during video capture,an adaptive brightness correction method based on the Gamma coefficient and conversion to the HSV color space is proposed.The brightness of the test image is effectively improved.The paper then studies and simulates the model of heart rate extraction from video sequences.A heart rate extraction model was established,and the selection of facial ROI regions,pulse wave signal extraction,frequency domain filtering,and heart rate extraction were studied.Two algorithms,peak detection and power spectral density analysis,are proposed to extract the heart rate from the pulse wave signal.Aiming at the shortcomings of the peak detection algorithm,a three-point mean method is proposed for peak detection of waveform signals,and the omission or misjudgment of peak points is greatly reduced.In this paper,the open source data set in Physio Net is used as a reference value,and the heart rate results detected by the algorithm are analyzed by Bland-Altman method and Kappa method.The results show that the peak detection algorithm is more accurate than the power spectral density algorithm,and the heart rate meets the measurement accuracy required by clinical medicine.Based on the algorithm simulation,the hardware design,software design and test plan of the heart rate detection system are given.The hardware includes a mini PC,camera and monitor.The system algorithm is divided into two parts: video micro motion amplification and video heart rate signal extraction.From the perspective of the user,the UI interface is designed to realize the function of the subject to measure the heart rate autonomously.The system also has functions such as a user’s personal database and video calls between users.At the end of the thesis,test cases was designed,and the function and performance of system was tested.In terms of functional testing,it is verified through data consistency analysis that the measurement accuracy of the system meets the requirements of clinical medicine.In terms of performance,the system has real-time performance and long-term stability.In summary,the system design meets the design goals.