Research On Non-contact Heart Rate Monitoring Technology Based On Near-infrared Images

In the medical field,heart rate is one of the key physiological index.If it is abnormal,some cardiovascular diseases might appear.Real-time and effectively monitoring heart rate might provide timely treatment for patients.However,due to the drawbacks of the traditional heart rate measurements,the new technique of non-contact monitoring appears.This method monitors the hear rate by capturing the color changing of human skin through the camera.However,this method might be affected by the change of illumination.In order to overcome it,a near-infrared-based heart rate detection method is proposed in this thesis.The main work is as follows:Firstly,the ordinary RGB camera-based heart rate detection system cannot work in the night scenarios.To overcome it,a near-infrared camera was used to capture the human face images.After a multi-task convolutional network was used to separate the face from the background,the nose area was chosen as Region of interest(ROI)for the heart rate information extraction by comparing with other areas on the face.To locate and track the ROI,here the facial landmark algorithm was implemented.Experimental results show that this method might accurately monitor heart rate in the dark scenarios.In addition,measurement distances and denoising algorithms were also discussed.The experimental results show that the proposed solution acheives good robustness and reliability.Secondly,to combat the problem of low signal-to-noise ratio in the single-channel near-infrared image,a multi-ROI fusion heart rate detecting method was designed depending on several areas such as nose,left and right cheeks,left and right forehead,chin and other regions.The ensemble empirical mode decomposition algorithm was used to decompose the original signal,and remove the high-frequency noise and low-frequency trend.Furthermore,an independent component analysis algorithm was used to remove the noise,and improve the signal-to-noise ratio.After that the peak signal-to-noise ratio was combined with Pearson coefficients to determine the heart rate.Experimental results showed that the near-infrared heart rate detection system based on multi-ROI fusion proposed in this thesis has an accuracy of 97.5% in the dark environment and 97% in the normal lighting environment.