Research On Indoor Fingerprint Localization Method Based On Channel State Information

The constant development and popularization of wireless telecommunication technology has resulted in an increasingly wide application field of Wi-Fi in daily life.Owing to low cost,broad signal transmission scope and strong adaptability,Wi-Fi is widely applied in different indoor positioning systems.No need of arrangement of special equipment gives Wi-Fi-based indoor positioning method a big advantage.According to different acquisition signals,Wi-Fi-based indoor positioning technology consists of indoor positioning technology based on Received Signal Strength(RSS)and that based on Channel State Information(CSI).However,under indoor environment,RSS is much likely to be disturbed by other signals as a kind of coarse-grained information,and is unable to provide high precision and reliability as impacted by indoor multipath effects.In the case where Wi-Fi signal of IEEE 802.11 n transmission protocol is used,CSI of Orthogonal Frequency Division Multiplexing(OFDM)subcarrier can be obtained by revising the wireless LAN driver.Coming from physical layer,CSI is able to show the channel characteristics and state between transmitter and receiver.The article mainly carries out researches as follows:(1)Characteristics of Wi-Fi-based finger print indoor positioning technology are compared with other different indoor positioning technologies.CSI finger print indoor positioning method by using Restricted Boltzmann Machine(RBM)is put forward,taking into account the actual application scenarios and demands.Upon data acquisition,CSI data is calibrated by RBM and finger print database is established.For on-line period,position is estimated by Naive Bayes Algorithm.According to the experiment results,with a small amount of finger print data,the method suggested by the article can provide a reliable positioning precision.(2)Concerning problems of insufficient data,single experiment environment and low robustness,a CSI finger print indoor positioning method by Discrete Hopfield Neural Network(DHNN)is put forward on the basis of the research of the first phase.In this method,finger print data is set as attractor and DHNN is introduced for convergence judgement,so as to estimate the location.Meanwhile,experiment environment in comparison is used to validate the steadiness and robustness of the said method.The final experiment results show that the said method presents a high positioning precision compared with similar methods.(3)Within a certain range,the larger amount is the data,the more detailed the finger print database is established,and the higher the positioning precision will be.However,the human and time cost of data acquisition will significantly increase to reduce the generality.Therefore,a CSI finger print indoor positioning method combining amplitude and phase position is put forward.In actual environment,data acquired by hardware equipment is analyzed and screened to select from the large amount of data the most representative of location characteristics,so as to establish more correct finger print database.As shown by the experiment results,compared with similar methods,the method suggested in the article can provide a reliable positioning precision with a strong robustness.