Research On The Method Of Non-contact Adaptive Robust Real-time Monitoring For Infant Heart Rate

Heart rate is an important indicator to evaluate the health status of infants.Heart rate monitoring is one of the important items for infant monitoring.A variety of reasons can cause changes in heart rate,such as fever,urination,defecation,intermittent asphyxia,infectious diseases(such as pneumonia,respiratory infections,diarrhea,electrolyte imbalance,etc.),organic heart disease(such as viral myocarditis,congenital Sexual heart disease),hypothyroidism,increased intracranial pressure,sepsis,etc.Therefore,it is necessary to monitor infants’ heart rate in real time intermittently or continuously so as to know infants’ heart rate information in time and discover the above-mentioned adverse symptoms as early as possible so that it can achieve the warning effect.At present,the monitoring of infants’ heart rate is mostly carried out by contact methods.Contact methods require specific physical contact between the subject and the device.It is cumbersome and inconvenient,and infants’ skin is delicate and not mature.Long-term use of contact heart rate monitoring can cause skin damage and physical discomfort,and impose a physiological burden on infants.Non-contact heart rate monitoring does not require sensors and cables to be in contact with the body.It has the advantages of non-invasive,convenient and comfortable in monitoring.However,the existing non-contact heart rate monitoring technology is not yet mature,the accuracy of heart rate monitoring and the robustness to facial expression changes and illumination changes are not good enough to meet the requirements of daily use.Due to the existing non-contact heart rate monitoring methods using a single fixed sensitive area for pulse wave extraction,and require the subject to keep the head still,but for infants,there are uncontrollable movements and facial expression changes on their heads,which result in poor accuracy and poor anti-interference ability to head movements and facial expression changes when used in infants’ heart rate monitoring.In addition,most of the existing non-contact methods do not consider the interference caused by illumination changes during heart rate monitoring.When there are changes in external illumination,it is easy to interfere with the monitoring results and affect the accuracy of heart rate monitoring.To solve the above problems,this paper introduces a joint face detection and alignment method to provide accurate facial landmarks quickly and robustly when there are head movements and facial expression changes,thereby weakening the motion artifacts caused by infants’ head movements and facial expressions changes;An adaptive selection method for non-fixed sensitive regions(local patches)is proposed to adaptively select and determine dimensionally stable local patches that contain only skin areas and are less affected by facial expression changes during monitoring,thereby improving the robustness of heart rate monitoring to infants’ head movements and facial expression changes;An improved reflected light intensity model is proposed.By introducing the reflected light intensity of the background region and taking the ratio of foreground and background signals to eliminate the time-varying light intensity term in the model,thereby weakening the interference caused by illumination changes and improving the robustness of infant heart rate monitoring.Firstly,in this paper,the existing methods and equipments of infant heart rate monitoring are reviewed from the perspective of contact and non-contact methods,and the research status of non-contact heart rate monitoring is elaborated from the aspects of interference of head movements and facial expression changes and interference of illumination changes.Secondly,the theoretical analysis of imaging photoplethysmography is introduced from three aspects: theoretical basis,working principle and realization condition,which lays a foundation for the extraction of pulse wave in the following paragraphs.Thirdly,in order to reduce the motion artifacts caused by the infants’ head movements and facial expression changes,it is proposed to use a joint face detection and alignment method to be able to quickly and robustly provide accurate facial landmarks when there are large head movements and facial expression changes;Fourthly,based on the obtained facial landmarks mentioned above,a method of adaptively selecting non-fixed-size local patches is proposed.The skin probability map(SPM)is constructed by the naive Bayesian model to filter out the non-skin local patches.The quartile distance is calculated by the box plot of the local patches,so as to adaptively select the size-stable local patches which contain only skin areas and are less affected by facial expression changes during monitoring.This adaptive selection method for local patches improve the signal-to-noise ratio of the pulse wave,thereby improving the robustness of heart rate monitoring to infants’ head movements and facial expression changes;Fifthly,in order to eliminate the interference caused by illumination changes,an improved reflected light intensity model is derived based on the traditional one.By introducing the reflected light intensity model of the background region and taking the ratio of foreground and background signals to eliminate the time-varying light intensity term in the model,so as to improve the robustness of infant heart rate real-time monitoring to illumination changes;Finally,the hardware and software platform of the system is built according to the requirements of the infant heart rate real-time monitoring system.The system is developed by the network surveillance camera + PC upper computer,and the human-computer interface of the upper computer is developed with Visual C++ as the development environment and debugging tools.The implementation of the algorithm is completed by combining OpenCV and Matlab.The existing method,which adopts single fixed sensitive area is compared with the proposed method based on the system mentioned above.The experimental results show that the proposed method has better accuracy and robustness.In addition,the system uses the infrared independent light source provided by the camera,which combined with the algorithm of this paper can realize the real-time infant heart rate monitoring in multi-scenarios including daytime and nighttime.The software platform of the system can be transplanted from PC to mobile phone,and the corresponding functions can be implement by mobile phone APP,which will increases the convenience of users.The non-contact infant heart rate adaptive and robust real-time monitoring method studied in this paper has good accuracy and robustness to infants’ head movements and facial expression changes and illumination changes.Even in the case of infants with large head movements and facial expression changes,along with conspicuous illumination changes,the method can still provide accurate and robust heart rate monitoring.The research in this paper provides practical and feasible ideas for family infant monitoring and medical patient monitoring,and it lays a foundation for the wide application of non-contact infant heart rate monitoring.So,this paper has high academic value and practical application value.