Research On Contactless Heart Rate Variability Feature Extraction Based On Real Application Scenarios

Heart Rate Variability(HRV)is an important indicator to evaluate autonomic nervous activity and cardiac intrinsic dynamics,and can be used in many fields such as medical health and emotion recognition.At present,HRV features are mostly obtained by contact method,but in some practical application scenarios,such as emotion recognition in customs travel inspection,it is difficult to obtain HRV features by contact method.Combined with the requirements of practical application scenarios,this paper studies the non-contact HRV feature extraction problem,and focuses on solving the efficiency of non-contact HRV feature extraction in practical application scenarios,as well as the impact of shaking and other factors on the accuracy of non-contact HRV feature extraction.Non-contact HRV feature extraction based on the principle of Imaging photoplethysmography(IPPG)takes a lot of time in image processing and signal processing,and the time efficiency of feature extraction often fails to meet the requirements of practical application scenarios such as automatic detection of passenger inspection channels.To solve this problem,this paper proposes a HRV feature extraction strategy based on concurrent processing and face tracking.According to the characteristics of non-contact HRV feature extraction key processing module,use multithreading technology through a Shared image queue to make concurrent execution data acquisition module and image processing module,signal processing and feature extraction module and the feature extraction process concurrent execution,and use the combination of face detection and face tracking method for facial image,optimize the image processing module,Improve HRV feature extraction efficiency.A large number of experiments in practical application scenarios show that,taking 30-second video collection as an example,the total time from video collection to HRV feature extraction is 32.63 s on average by adopting the strategy proposed in this paper,which is 52.31 s shorter than that without adopting the method in this paper,and can meet the time requirements of practical application scenarios.Aiming at the problem of incomplete region of Interest(ROI)in face tracking extraction due to the wobble of the detector,which affects the accuracy of HRV feature extraction,this paper proposes an affine transform-based ROI correction method,which realizes ROI correction through affine transformation of the tracking results.In order to solve the problem of automatic checker switching in practical applications,this paper proposes an automatic checker switching method based on ROI determination.In a large number of application tests,the comparison experiment with the HRV feature extraction by contact method shows that the proposed method can achieve ROI correction and automatic switch of the detector,which not only improves the accuracy of non-contact HRV feature extraction,but also does not affect the efficiency of HRV feature extraction.In this paper,a non-contact HRV feature extraction and emotion recognition system is designed and implemented based on the research results of this paper.The system includes a collection terminal and a management terminal,which can extract the HRV characteristics of customs clearance personnel in multiple passenger inspection channels and identify whether customs clearance personnel have abnormal emotions,so as to assist customs staff in screening suspicious persons and criminals.At present,the system has been put into trial operation in a customs.