Research On Device-Free Localization Method Based On Wi-Fi Probabilistic Imaging Model

In recent years,with the rapid development of new-generation information technologies such as the Internet of Things and artificial intelligence,wireless sensing technology has gradually become the focus of attention in the field of production and life.Utilizing the non-contact and wall-penetrability properties of radio signals,wireless sensing can realize sensing functions without the target carrying any wearable sensor equipment.At present,wireless sensing has broad application prospects in intelligent sensing applications such as smart home,smart medical treatment,and environmental modeling.Wi-Fi imaging is one of the important technologies of wireless sensing.It can use the existing Wi-Fi devices in the observation area to perceive and image the monitored target,and then realize extended applications such as device-free localization and human behavior recognition.Now Wi-Fi imaging technology has gradually become a research topic of general importance in the academic and business communities.Wi-Fi imaging requires an imaging model to form an image mapping of spatial points and pixel points.However,the existing Wi-Fi imaging researches still lacks universally applicable imaging models.The main imaging models such as Radio Tomographic Imaging(RTI)and Angle-of-Arrival-based scene imaging models have high requirements for signal transceivers,which are usually inefficient and difficult to be widely deployed.Meanwhile,due to the limited number of antennas and bandwidth resources of Wi-Fi,the existing Wi-Fi imaging systems generally have low resolution.Although space diversity or frequency diversity technology is used to increase imaging signal sources,it is still difficult to compensate for the imaging difficulties caused by few data samples.Therefore,it is a research challenge in the field of Wi-Fi imaging to build an effective and applicable Wi-Fi imaging model and improve the imaging resolution.To address the lack of effective physical models for the existing Wi-Fi imaging technologies,this paper proposes a point probability imaging model based on Channel State Information(CSI),and constructs a Point Spread Function(PSF)based on Gaussian function for Wi-Fi imaging,and proposes a Gaussian Mixture Model(GMM)based Wi-Fi sparse image reconstruction algorithm.In this paper,we analyze the CSI in commercial off-the-shelf(COTS)Wi-Fi devices,explore the mechanism of environmental information affecting and modulating Wi-Fi signals,innovatively model the influence of spatial objects on signals as point impulse response,and use this as a starting point to establish a new Wi-Fi imaging model based on grid-point probability.In order to make the model more generalizable and robust,the model is generalized to a practical scenario.In addition,a PSF-based image degradation model is introduced to cope with the blurring problem in the imaging process,and a GMM is used to model the PSF to achieve a better fit.Finally,the target image is reconstructed by Bayesian estimation based on Compressed Sensing(CS).Through the analysis and verification of simulation and measured data,the proposed model is able to recover the planar image of the target from less than 10 antenna signals,and the average localization error in indoor environment can be reduced to 1.047 m.The PSNR of the recovered image reaches 39 d B.Compared with other existing Wi-Fi imaging researches,the final imaging results of this paper have relatively small target localization error and high images reconstruction quality,which demonstrates that the proposed imaging model and image reconstruction algorithm have strong robustness.