Research On Non-contact Respiration Frequency And State Recognition Based On Radio Frequency Signals

Breathing is one of the important life activities of the human body,and it can provide important information related to human health.Accurate detection of respiratory rate and respiratory state helps prevent and identify lesions in the lungs,respiratory tract,and bronchus.Clinically,most of the commonly used breathing detection systems are contact type,and the measurement equipment needs to be in direct contact with the human body to obtain information related to breathing.However,this contact-type detection method has certain compulsiveness,which usually makes users feel uncomfortable.In addition,these contact devices are not only difficult to operate,but also have relatively high maintenance costs,which is not suitable for large-scale promotion and use.Therefore,it is of great significance to study non-contact breath detection technology with low cost and no trauma to users.Under this background,this topic focuses on contactless breathing frequency estimation and breathing state recognition.The measured received signal strength(RSS)information and channel state information(Channel State Information(CSI))are used to achieve For accurate detection of target breathing,the main work of the thesis includes the following aspects:(1)In view of the shortcomings of the contact breathing detection system,this article uses related hardware platforms to build a 2.4GHz band breath measurement system based on Zigbee,a 915 MHz band breath measurement system based on Si4463,and a 5GHz band breath measurement using Intel 5300 network card system.Using three frequency band non-contact respiration measurement system to achieve low-cost respiratory data collection.(2)For the estimation of breathing frequency,this article uses the breathing data collected in the above three frequency bands to estimate the breathing frequency of the measured target.First,the independent component analysis(ICA)method is used to preprocess the collected data to reduce the influence of noise,and then the compressed sensing frequency estimation method is used to obtain a more accurate respiration rate.On this basis,in view of the adverse effects of body motion on breath detection,the fluctuation index feature is used to determine the period of motion interference to improve the accuracy of breath detection.(3)By analyzing the signal changes of multiple sub-carriers(or multiple links),two-dimensional respiratory characteristics are given,and on this basis,two respiratory state recognition methods are proposed.The first method uses the texture features of the two-dimensional respiratory image to determine the similarity and determine the corresponding respiratory state;the second method uses a Recurrent Neural Network(RNN)to collect the different respiratory state data Classification recognition operation.Finally,the research work of this paper is summarized,and the shortcomings in the existing work are summarized,and suggestions for improvement are also put forward to improve in the future research.