Research And Implementation Of Non-contact Apnea Detection Algorithm

Sleep apnea syndrome is a common type of respiratory related disease,which not only affects the daily life of patients,but also causes high blood pressure,stroke,diabetes and other diseases.PSG is currently used in medicine as the gold standard for sleep apnea syndrome testing.However,PSG check is expensive and cumbersome to operate,requiring a large number of sensors to directly connect to the body and interfere with normal sleep.Therefore,this paper proposes a low-cost,easy-to-operate,sensorfree detection method for the early screening of sleep apnea syndrome,which is of great significance and value to patients,medical and health care institutions.This paper refers to the current detection method of sleep apnea syndrome,and proposes a detection algorithm based on PVDF piezoelectric film sensor signal for sleep apnea syndrome.The algorithm is non-contact in the acquisition mode,which greatly reduces the burden on the tester.Using the original piezoelectric film signal,data enhancement,sliding window segmentation,and related data sets were made;a 6-layer one-dimensional convolutional neural network was used to construct the training model,and the input signal selection and model optimize were finally used.The apnea detection model has an accuracy rate of 92.76%.In addition,this paper discusses the possibility of using the sensor signals of the chest to achieve categories of apnea recognition.By constructing a simple one-dimensional convolutional neural network model and comparing it with the one-dimensional convolutional neural network for apnea detection,the parameters of the final apnea detection were 77.13% accurate by parameter optimize.After that,the model of apnea detection and category identification was comprehensively tested,and the accuracy of 74.67% was obtained.Finally,this paper also designed and developed an apnea detection system for the apnea detection algorithm.The system encapsulates the neural network model algorithm,and uses the micro-service architecture to invoke the algorithm through the API interface to realize a simple and extensible operation B/S system.