Research On Heart Rate Measurement Algorithm Based On Face Video

As one of the four physiological indicators of human beings,on one hand,heart rate can be used to predict the risk of cardiovascular disease,on the other hand,it can be used as an important criterion for sleep quality assessment.Therefore,daily heart rate detection is essential for human health.At present,the heart rate measuring instruments widely used require direct contact with the skin of the subject,which is complicated to operate and expensive.Long-term contact is likely to cause discomfort to the subject,and is not suitable for burn patients with damaged skin and delicate newborns.Based on the principle of imaging Photoplethysmography,this thesis extracts heart rate information from human face video under natural light by face detection,blind source separation,adaptive filter and other algorithms.This method has the characteristics of simple operation and low cost,and can realize continuous non-contact heart rate measurement of the subject.At present,the heart rate measurement algorithm based on face video has general adaptability to motion artifact and light variation,and the algorithm has not been integrated into the system.Based on this,the main work of this thesis includes:(1)Extract source heart rate signal based on face video.The adaptability of different face detection algorithms is evaluated in different scenarios,and the comparison charts of accuracy and speed are given.At the same time,different algorithms of the region of interest are tested.Finally,the region is obtained from the video frame.The Blind Source Separation is used to separate the source estimation signals from the three-channel signal,and the correlation coefficient is analyzed to obtain the source heart rate signal.(2)Calculate heart rate.Analyze the type of noise and the corresponding algorithm in heart rate signal.The source heart rate signal and the background area signal are used for adaptive filter to remove the interference of illumination variation and the filter parameters are quantitatively given.Bandpass filter is used to remove the power frequency interference and myoelectric noise.Finally,the time domain and frequency domain heart rate calculation methods are given respectively and the error comparison analysis is carried out.(3)Based on the heart rate detection algorithm in this thesis,offline and real-time heart rate detection systems are implemented.The system architecture of offline heart rate detection is given,the processing module of high-concurrency flow is explained in detail,the algorithm flow chart of offline data calculation is drawn,and the front-end interface of data display is given.At the same time,the real-time heart rate detection system is implemented,the algorithm principle and flow chart of the real-time heart rate detection system are given.(4)Evaluate the accuracy and adaptability of heart rate calculation in this thesis.The Bland-altman diagrams are used to prove the strong consistency and linear correlation between the scheme and the standard values in the scenarios with motion artifacts and illumination variation respectively.Comparing the average errors of different algorithms,the results show that the performance of this scheme is better than other algorithms.Finally,the experiment is carried out under different illumination conditions,and the results show that the scheme is adaptable to different illumination scenarios.