Research On Respiratory Belt Data Analysis And Disease Feature Extraction Method

The physiological signals of the human body,such as breathing and heart rate,can reflect the health status of the human body,and as people pay more and more attention to the health status of the body,the research and analysis of the physiological signal data becomes more important.Human respiratory signal(Respiratory,RSP)is a basic physiological signal of the human body.Its intensity,frequency and other information can largely reflect other human functions such as heart and lung function characteristics,and can reflect the respiratory system itself such as the lungs,bronchial and other parts of the disease.This paper analyzes the data of respiratory signals,aims to explore a method for positive and abnormal diagnosis of respiratory signals based on artificial intelligence and machine learning technology.At present,the analysis methods for respiratory data are not mature enough,mainly focusing on the acquisition and preprocessing of the original respiratory data,but the existing acquisition methods such as impedance detection and extraction from ECG signals are not stable enough and there are problems such as irritation to the human body.Furthermore,it is not clear which features of respiratory data can effectively classify the respiratory system.In response to the above problems,the main work and contributions of this article can be summarized as the following three points:1.Use the breathing belt,a portable,non-invasive device as a respiratory data acquisition tool,and propose a method to remove noise interference from respiratory signals.After removing high-frequency noise using a Butterworth low-pass filter,through data translation and window sliding filtering and other methods to remove the baseline drift interference,then get a clean breath signal to provide a data basis for the next work.2.Through the analysis of the characteristics of respiratory data,17 types of characteristics that may have an impact on the classification of respiratory signals are proposed,and these characteristics are discussed using different statistical analysis methods such as sensitivity and specificity,and the highest discrimination is selected.Several types of characteristics are used as the basis for subsequent experimental research.3.Propose a classification method based on DAE's Long Short-Term Neural Network(LSTM)combined with manual feature extraction.The DAE structure is used to extract the unsupervised feature representation,and an expert mechanism is introduced-active learning.After artificially labeling the data,the classification model is retrained,and finally combined with the manually selected features which has good effects.Experimental results show that the proposed method can achieve better classification results than traditional classification methods.