Passive Indoor Location Method Based On CSI

In the current LBS field,electromagnetic signals such as WIFI can be used in the indoor position service as a hotspot in the marketing.However,the complex structure of indoor environment will cause problems such as shielding interference in wireless signals,which resulting in the multipath effect of electromagnetic signals,thus seriously affecting the accuracy of indoor positioning.In view of the problems existing in WIFI indoor positioning as above,nowadays,basing on OFDM technology,this paper chooses CSI signals which collected from WIFI physical layer,and take it as the positioning information source instead of RSSI which is the traditional one from Mac layer.The reason is that CSI has more abundant characteristic data,so it can performance more significant differences in the mapping relation among different indoor positions,which indicates that CSI has higher stability as a positioning information source,so CSI can support passive indoor positioning better.To sum up,based on the general principle of existing wireless signal fingerprint positioning,this paper analyzes and studies the influence of using CSI as the indoor positioning information source in both two stages of fingerprint location method,which are offline collection stage and online matching stage.First,in the off-line sampling preprocessing stage: in order to solve the problem that mutual disturbance exists among the adjacent sub-channels of CSI,the hybrid subchannel differential method is adopted to generate the positioning fingerprint.The specific method is to suppress the mutual interference of adjacent sub-channels by means of fixed distance difference,so as to compensate for the data lost in fixed distance difference by using ordinary difference fingerprint as supplement,this method can also maintain the integrity of origin data set.In addition,the performance of the fingerprint construction algorithm was evaluated from the aspects of antenna numbers,algorithm complexity,fingerprint feature selection,etc.Further more,a hypothesis of wireless signal distribution law was proposed at this part,and an independent experiment was designed to verify it.Secondly,in online matching phase:The Hidden Markow Model was improved and optimized,and mathematical modeling was carried out according to the indoor fingerprint location method.In view of the problem that the random initial values of the prediction model parameters of BW in HMM are easily trapped in local optimization,this study proposes a PSOHMM algorithm.The fusion algorithm introduced PSO in the initial value selection of BW algorithm to make the result closer to the global optimum.Meanwhile,in order to adapt the constraint of statistical data in the HMM algorithm model,re-standardization and re-mapping mechanism were added to the data exit of PSO algorithm.Finally,on the basis of the experiment results of this article,the results of offline fingerprint preprocessing show that the effect of offline fingerprint localization that based on CSI information is much better than the one based on traditional RSSI,which can reach an average cumulative error of 1.16 m.At the same time,Experiments in the online matching stage show that the positioning effect of PSO-HMM fusion algorithm is slightly better than other mainstream fingerprint matching algorithms such as SVM algorithm and the ordinary HMM algorithm.The result take both accuracy and timeliness into account,with the fingerprint matching rate reaching position accuracy of 95% and the average positioning time at 17.8s.