Research And Application Of Breathing And Heart Rate Detection Based On Deep Learning

As two important internal organs of the human body,the heart provides sufficient power source for blood circulation and the lungs provide a strong guarantee for respiratory circulation,both of which are important components of the body.However,with the increasing prevalence of cardiovascular and respiratory diseases,these diseases have begun to seriously endanger human life and health safety,so people should pay more attention to the monitoring and prevention of cardiovascular and respiratory diseases.This paper presents the basic theoretical research and application of a non-contact piezoelectric ceramic monitoring device and deep learning technology.First,the current research status of respiratory heart rate acquisition methods at home and abroad and the shortcomings of the methods are analyzed and summarized;then,the basic theories related to deep learning are introduced and applied to heart rate respiratory analysis of piezoelectric signals for the first time,and multidimensional processing of piezoelectric signals is proposed to optimize the prediction performance of the model.The study constructs a point-to-point mapping model from piezoelectric signal to respiratory heart rate data;then,using the obtained respiratory heart rate data,the application study is carried out and a sleep staging algorithm based on respiratory,heart rate and body movement data is designed to realize the analysis of sleep quality;finally,the prediction model and staging algorithm are integrated to develop a health monitoring platform based on piezoelectric ceramic sensors,which facilitates users to accurately control personal health information.Finally,we integrated the prediction model and staging algorithm to develop a health monitoring platform based on piezoelectric ceramic sensors to facilitate users’ accurate control of personal health information.Based on the above theory,this paper conducts experiments to build prediction models using the original data and the data after multidimensional processing,and the results demonstrate the feasibility and advantages of multidimensional processing of piezoelectric data.And comparison experiments were conducted using RNN,ARIMA classical prediction network and LSTM network,and the results showed that the prediction effect of LSTM prediction network based on time-frequency feature extraction has better overall performance in four evaluation indexes.On the other hand,the error of the sleep stagingalgorithm was kept below 10%,which met the higher requirement.Finally,the overall function of the health system was tested,and the results proved that the health information monitoring system has practical functions and accurate data,which can meet the daily monitoring needs of users and has great practical significance.