| According to the report of China Cardiovascular Disease Report 2017(Summary)cardiovascular mortality accounts for more than 40%of the deaths of residents in China,and the number of deaths has increased year by year,indicating that the situation of cardio-vascular disease prevention and treatment is grim.Heart rate is one of the important indicators of cardiovascular health,so daily monitoring data of heart rate and its trend is of great reference value for prevention and early diagnosis of cardiovascular disease.As residents'health awareness increases,the need for daily health monitoring is growing as well.Contact method is mostly adopted to measure the heart rate in clinical medicine,which probably makes patient's movement impaired,meanwhile,long-term monitoring is easy to cause skin discomfort,unsuitable for daily monitoring.Non-contact and non-somatosensory heart rate detection technology with great appli-cation potential is becoming a research hotspot at home and abroad in the field of daily health monitoring.In this paper,we study an imaging heart rate detection method and design an imaging real-time heart rate monitoring system based on Qt platform and C++language,which has various working modes.The work of the thesis lays a great foundation for the realization of heart rate detection and long-term monitoring in the home and office environment.The specific work is as follows:The imaging heart rate detection method described in this paper is based on the Beer-Lambert law and the image correlation principle.Firstly,we use the Pearson correlation coefficient to quantitatively analyze the ROI pixel matrix of the frame to obtain the correlation coefficient queue.Then,we decompose the signal queue by wavelet and transform the reconstructed signal queue into the frequency domain by DFT.Through the spectrum analysis of signals,the heart rate values with high accuracy and precision are finally obtained with the processing of interpolation and non-heart rate frequency region filtering.We collect the skin area images of several cardiac cycles in real time through the camera when acquiring the source image,and different color channels will be chosed to quantitatively measure the ROI similarity between adjacent frames according to different working modes.The system can implement the ROI extraction based on face feature point location,which can in some way get rid of the activity constraints of the testee.Meanwhile,a dynamic queue suitable for real-time data stream updating is designed to meet the real-time heart rate monitoring requirements.In terms of the filtering,we use the wavelet filtering to filter the signal denosing,ulteriorly determin the wavelet base with good filtering effect and the corresponding vanishing moment coefficient experimentally,with the optimal maximum signal recon-struction component number determined as well.We design a direct non-heart rate zone frequency filtering method based on the queue,and then extract the peak value of the spectrum at 0.8 Hz-3.7 Hz to obtain the heart rate,displaying and updating the key data in real time and dynamically.In addition,we build a light path for defocusing obtaining laser speckle patterns for the amplification of microvibration on the skin surface.The system has realized a variety of measurement modes such as speckle pattern,fingertip pressing,mouse custom,and fully automated,which are suitable for different precision and measurement scenarios.In the aspect of measurement analysis,experiments are carried out to verify the accuracy and robustness of the system in four working modes.The real-time correlation coefficient waveform,spectrum peak position and heart rate data are recorded in 5 minutes for subsequent analysis.The experiments show that the system can adapt to the heart rate monitoring needs well,with the heart rate data basically consistent with the pulse number and its fluctuation very samll.Finally,the factors that may cause errors in all aspects of the system are analyzed,and suggestions for improvement are put forward. |
PDF Full Text Request