Indoor Locating Method Research Based On RSS And CSI Fingerprint

With the proliferation of WiFi and mobile devices,WiFi based indoor positioning is being watched by more and more researchers.Receiving signal strength indicator(RSS)as a mainstream solution is often used for range-based positioning system and fingerprint positioning system.However,RSS is often affected by multiple size effects and noise signals,and its location performance is not stable.In recent years,commercial WiFi devices,such as the Intel 5300 wireless network card,support access to the physical layer’s channel status information(CSI).CSI is an indicator of a more fine-grained characterization of signal characteristics than RSS.Compared to RSS,CSI analyses the characteristics of multiple subcarrier signals to avoid the effects of multipath effect and noise.The CSI has opened up new Spaces for WiFi based location-based technology,and is being watched by researchers.But because of the heterogeneous wireless network environment,the CSI is vulnerable to certain subcarrier signals.And most of the methods based on the CSI fingerprint don’t take advantage of the valid information in RSS to accelerate the location.And most importantly,because the 802.11 n standard AP deployment is still sparse,most of the AP is still a mix of 802.11 b/g/n,and for this class AP its CSI information is not available for the time being.Due to the indoor environment is vulnerable to the influence of the multi-path effect,coarse-grained received signal strength(RSS)will be affected by the larger,so the traditional positioning accuracy is not high based on RSS.While using fine-grained physical channel state information(CSI)can to a certain extent,to avoid the impact of the multipath effect,but the general needs to place of indoor positioning applications(store)its indoor environment is complex,unable to effectively estimate the direct use of CSI information AP and pending the distance between the target.Through the experiment we choose the related AP,and then we collect the RSS and CSI experimental data of the experimental environment data,and the data to do the cleaning,eliminating the spike of RSS,in addition to the serious interference of the CSI data,and using the SVD method to do the related data recovery work.After we have our experimental environment can be divided into small squares area of initial position of the first preliminary range positioning,using RSS information behind CSI information for precise positioning.For the CSI information,this article first identifies the CSI information that is disturbed by heterogeneous signals.Because of the complexity of the indoor positioning applications under the CSI data with high dimensions,this article does not use CSI information similarity comparison or feature extraction under linear space,but the depth using neural network to CSI characteristics of black box modeling,to avoid explicit information extraction of available CSI characteristics on the basis of fully express between CSI fingerprint difference between different locations,in particular in this paper,the different position training different neural network and using the corresponding weights of neural network as a "fingerprint",pending a target the CSI fingerprints as a different position of the corresponding depth of the neural network input,by comparing the different location of the depth of the neural network output and the backlog of target CSI fingerprint gap,using bayesian inference finally decided its precise location.Core algorithm in this paper,the main structure is mainly composed of two parts: first,the use of multiple single AP RSS information preliminarily locate respectively,finally voted to target roughly the area.This method is mainly divided into offline fingerprint database construction stage,the stage of online RSS fingerprints,RSS fingerprints weights allocation strategy,as well as the largest connected subgraph in regional choice.Then,based on the CSI data,the black box was modeled using a deep neural network algorithm,and then used bayesian algorithms to pinpoint the location.This method can be divided into the architecture of DNFN,each pending area and the depth of the AP of the neural network construction,and finally using bayesian algorithm selected the most suitable match fingerprint region.Based on the above reasons,this paper carried out the research on the indoor location method based on RSS and CSI hybrid fingerprint.