Research On Exercise Heart Rate Detection Algorithm Based On RPP

Heart rate,as one of the most important clinical indicators of a patient’s physical condition,is a parameter used to measure the beating capacity of the heart.Heart rate detection has a wide range of applications in the fields of medical care,fitness sports,and emotion recognition.At present,mature heart rate detection products are mainly contact measurement by electrocardiograph and photoelectric volumetric pulse wave tracing（Photo Plethysmo Graphy,PPG）,but the disadvantages of high hardware requirements and more complex operation make the emergence of non-contact heart rate detection inevitable.Conventional non-contact heart rate detection is mainly achieved using rPPG（Remote Plethysmo Graphy,rPPG）signals.The skin color signal is obtained by spatially averaging the pixels in the ROI region.The rPPG signal extraction algorithm is used to realize the rPPG signal extraction,and then the heart rate calculation is realized by the peak detection method based on time domain features and the fast Fourier transform method based on frequency domain features.In this paper,we study and improve the traditional non-contact heart rate detection based on rPPG signal and proposed an rPPG heart rate detection algorithm based on the improved HMM algorithm.In the face detection module,the speed and accuracy of three different face detection algorithms used in the heart rate detection framework are compared in this paper.In the rPPG signal extraction module,the R,G and B channel chromaticity averaging based method,CHROM algorithm and POS（plane-orthogonal-to-skin）algorithm are implemented and compared in this paper,and the advantages of POS algorithm are illustrated.In addition,in the preprocessing of rPPG signal,this paper adds NLMS adaptive filter to filter out the noise brought by periodic motion signals to rPPG signal,which enhances the robustness of the algorithm in motion scenes.In terms of heart rate calculation,this paper proposes a frequency domain tracking approach based on the time-frequency map to realize the heart rate calculation.In terms of frequency domain tracking,this paper uses particle filtering model and hidden Markov model to implement and fit the heart rate variation amount of the subject in a large amount of video data in the database,based on which an improved HMM（Hidden Markov Model,HMM）algorithm is proposed.In this paper,experiments on the rPPG heart rate detection algorithm based on the improved HMM algorithm are conducted using the self-filmed database,and each experimental video includes three states:stationary,slow motion and vigorous motion.The experimental results of our own dataset are as follows:the mean error(₀))is 4.6636,the standard deviationis 7.1776,the root mean square erroris 7.1776,the Pearson correlation coefficientis 0.8719 and the?（78）≤5 is 0.8440.The experimental results of V4V dataset are as follows:the mean error(₀))is 3.2640,the standard deviationis 5.1890,the root mean square erroris 5.1890,the Pearson correlation coefficientis 0.9860 and the?（78）≤5 is 0.8910.The experimental results show that the heart rate detection algorithm proposed in this paper can achieve more accurate heart rate detection in sports scenarios.