Research On The Technology Of Human Respiration Detection And Classification Based On CSI

In all kinds of human daily activities,respiratory activity parameters are important indicators of vital signs information.At present,the detection methods used in the respiratory detection system are mainly divided into contact and non-contact.The contact method will generally affect the comfort of users,and the use of scenarios has limitations.Among the non-contact methods,the electromagnetic wave detection method has relatively good performance,but the traditional electromagnetic wave detection method relies on complex and expensive hardware equipment,which limits its practicability in many applications.Aiming at these problems,a human respiratory electromagnetic wave detection method based on Channel State Information(CSI)technology is proposed,which is cheap and has good accuracy and robustness.In addition,based on the CSI data,the classification problem of respiratory motion was studied.In order to solve the problems of traditional breathing detection methods,firstly,the feasibility of CSI technology is analyzed;secondly,a breathing information extraction model is established;lastly,experiments are carried out to collect the CSI data of human breathing and the data of strain breathing sensor as a control,which provides the theoretical basis and number for the detection of human breathing information and the classification of breathing movement.According to sources.For the detection of human respiratory information,firstly,according to the characteristics of CSI data,several traditional filtering algorithms,such as Fourier transform,Gabor transform,wavelet transform,Butterworth filter,Chebyshev filter and Kalman filter,are compared.Then,a suitable threshold filtering method is proposed to improve the wavelet transform.The relative error of the extracted breathing information is less than 7%.At the same time,it has good robustness and good detection effect.After comparing the results of five classification algorithms,such as J48 decision tree,K-nearest neighbor,support vector machine,BP neural network and Naive Bayes,the BP neural network is selected as the classification algorithm of human respiratory movement.A new preprocessing method is proposed for this algorithm,which is to obtain the covariance value of each layer of wavelet coefficients as the eigenvalue of frequency domain to make the preprocessing.After that,the eigenvalues are more differentiated,and the classification accuracy of this algorithm reaches 99.22%.Then,the neural network is adjusted to optimize the classification model,and the accuracy reaches 99.78%.