Research On Passive Target Motion Detection And Localization And Tracking Technology Based On WiFi Channel State Information

Passive target sensing technology based on wireless signals,with its non-contact and non-interventional sensing advantages,is widely used in various intelligent applications,such as smart security and home automation.In these application scenarios,the person to be sensed does not need to carry any device.Meanwhile,with the rapid development of Integrated Sensing and Communication,the use of wireless communication networks for passive target sensing services has become a mainstream trend.In particular,owing to the ubiquitous deployment of WiFi networks in most office and residential buildings,the WiFi-based solution has become one of the most promising wireless sensing schemes.As essential parts of wireless sensing applications,passive target motion detection,localization and tracking are also basic works of complex sensing functions.So far,many WiFi-based passive target motion detection and localization and tracking methods have been proposed.However,in practice,there are still many challenges to those methods.For example,signal models or signal characteristics depend on background environments,passive sensing systems have low sensing accuracy and poor robustness with limited wireless resources,and passive sensing function fails when the locations of the sensing devices change.So,this dissertation studies the single-link-based passive target motion detection and localization and tracking technology to address these issues.The main research and innovations of this dissertation are as follows:(1)This dissertation proposes a distribution-difference-based passive target motion detection algorithm to deal with the problem that signal features and detection methods are sensitive to changes in the background environment.Firstly,the channel temporal autocorrelation under static and motion states is analyzed.The autocorrelation coefficients of the power of Channel State Information(CSI)in the two states are then deduced.So,an environmental feature is designed to describe the distributions of the autocorrelation coefficients of the two states.Secondly,a clustering-vote-based sensitive sensor selection scheme is proposed to improve motion detection accuracy by fully utilizing the available carriers on all antennas.Thirdly,to avoid the worst effects of the background environment change,the distribution of environmental features under the silent state is regarded as the template profile for motion detection,independent of the background environment.With the template profile,a distribution-difference-based anomaly index calculation method is proposed.Then,a motion detection strategy based on double thresholds and hysteresis tracking is designed to enable reliable and robust motion detection.Finally,the experimental results show that the proposed algorithm can achieve a motion detection rate of more than 96.2% with one single WiFi link.(2)Aiming at the problem of low localization accuracy caused by complex and imperfect phase error calibration methods,a single-link-based passive target localization and tracking algorithm without phase error calibration is proposed.Firstly,to eliminate the interference of phase errors on parameter estimation,a multi-dimensional signal model based on CSI power is derived,which brings ambiguity in the sign of multidimensional parameters.Secondly,an Iterative Path Refinement-Space-Alternating Generalized Expectation-maximization(IPR-SAGE)method is used to estimate multidimensional parameters,combining CSI data of multi-frequency,multi-antenna and multi-package to improve signal resolvability.Then,a cross-correlation-based motion direction estimation algorithm is studied to eliminate the ambiguity of the signs of parameters.Thirdly,an adaptive Den Stream clustering algorithm is proposed to identify valid target paths and filter out redundant clutter paths.Fourth,interacting multiple models joint probability data association algorithm is used to track the passive human,which uses all valid echoes and potential target motion models to improve localization and tracking accuracy.Finally,the experimental results show that the proposed algorithm can achieve good localization and tracking performance without phase error calibration.The localization accuracy of a single user can reach 0.69 m when using one WiFi link.(3)Aiming at application scenarios where sensing devices can be moved at will and their locations are unknown,a novel passive target localization and tracking algorithm is proposed using one pair of commodity WiFi devices with unknown locations.Firstly,a multipath triangulation localization model is built to locate sensing devices and one passive human,and it is proved that the locations of the devices and the person can be estimated simultaneously in a passive manner.Then,a two-dimensional multiple signal classification algorithm is adopted to estimate the angular directions of all propagation paths,and the direct path is screened out by its path features.Secondly,due to the ambiguity in the sign of parameters,the proposed path parameter augmentation algorithm utilizes parameter flipping processing to obtain more reflection paths.According to the clustering characteristics of target reflection paths in different parameter domains,a confidence-aware-based parameter identification method for target paths is proposed to combat environmental clutter and noise.Thirdly,since the transmitter’s location is unchanged over a short period of time,a kernel density estimation method is used to estimate the relative location of the transmitter.Then,with the transmitter’s location estimate,a Gaussian Sum-Extended Kalman Filter(GS-EKF)algorithm is used to track the passive human with one single WiFi link.Finally,the experimental results show that the proposed algorithm can achieve a median localization error of 0.8 m for the passive human and a median localization error of 0.54 m for the transmitter.