Research On Contactless Respiration Detection In WiFi Wireless Scenarios

With the widespread deployment of WiFi wireless network infrastructure,more and more attention has been paid to the field of intelligent wireless perception,which uses advanced signal processing and artificial intelligence technology to sense,detect,track and identify the environment and human activity information carried by WiFi signals.And it has spawned a series of sensorless and non-intrusive intelligent sensing applications.In the field of medical and health care,detection of abnormal respiration in special population is a sensitive life indicator.Therefore,the research of non-contact breath detection based on WiFi signals is gaining increasing attention.However,due to the complexity of wireless signal propagation effects,we are faced with the challenges of extracting effective signal features,establishing fine space-time models,and detecting multi-person breathing scenarios.In this paper,commercial WiFi equipment is used to build indoor breath detection experimental environment,and the breath detection problems of different signal characteristics are studied from the perspective of wireless channel model.Firstly,the theoretical basis of intelligent wireless perception is introduced.In order to extract environment or human activity information from multipath signals effectively,time reversal(TR)technology and diversity technology are used to treat multipath signals as virtual antennas to provide high-resolution Space-time resonance characteristics.Secondly,according to the RSSI characteristics of WiFi signals,the quantitative relationship between the diffraction gain and the human thoracic displacement is proposed by using the Fresnel Zone knife-edge diffraction mode,and the influence of the human body position on respiratory detection is demonstrated.Experiments show that the accuracy of estimating respiratory rate at the appropriate location is 93.8%.Finally,according to the fine-grained CSI characteristics of WiFi signals,the CSI signal model is decoupled by Jacobi-Anger expansion and projected into the time Reversal Resonating Strength(TRRS)feature space to amplify the fine cyclic changes of CSI caused by breathing.Root-MUSIC algorithm is used to obtain candidate respiratory rates from TRRS feature space,and Affinity Propagation(AP)clustering,probability allocation,and cluster merging are used to estimate the respiratory rates of multiple people and to detect the number of people at the same time.The method is validated in two typical indoor scenarios and the results show that the average accuracy of single breath rate estimation is 96.6% and 94.3%,respectively.In multi-person scenarios,94.5% average estimation accuracy and the number of people with an error of less than one person can be obtained for 1 to 6 people.