Research On Single Point Localization Algorithm Based On CSI Via WLAN

The indoor localization system based on Wi-Fi network can take into account the localization accuracy and deployment cost,which has become a hot spot of indoor localization research.Traditional mainstream schemes usually use Received Signal Strength(RSS)and Received Signal Strength Indication(RSSI)as parameters to implement indoor localization system.However,RSS and RSSI are coarse-grained values of the MAC(Media Access Control)layer,which can easily fluctuate due to environmental factors.This makes it difficult for the localization system based on RSS/RSSI to achieve better localization accuracy.In recent years,researchers have been able to extract the fine-grained Channel State Information(CSI)of the Physical Layer(PHY)from some common commercial network cards by modifying the firmware,which opens up a new research space for indoor localization technology based on Wi-Fi network.In addition,in a small indoor environment such as a conference room,a personal home,and a small shop,there is usually only one wireless Access Point(AP)node,so the research on the single AP indoor localization system based on the Wi-Fi network has a wider application prospect.According to the above reasons,this paper takes CSI as the localization parameter to study single AP indoor localization based on ranging and fingerprint.The main research results are as follows:Firstly,this paper proposes an indoor localization scheme based on the combination of ranging and fingerprinting by means of annular sampling mode.In this way,the overall fingerprint set can be divided into several small fingerprint sets,and the fingerprint processing overhead in the fingerprint-based scheme is reduced by the ranging scheme,and the shortcoming of the accuracy based on the ranging scheme is compensated by the fingerprint scheme.In addition,by combining the ranging scheme and the fingerprint scheme,the requirement of the number of APs can be reduced,so that the scheme can be implemented via a single AP.Secondly,this paper uses the Intel 5300 network card and the CSI tool to capture raw CSI data from two typical indoor environments.Since the original CSI data obtained from the actual network card contains both the state information of the external wireless channel and the internal circuit state information,the purpose of this paper,however,is to utilize the CSI extracted in the commercial network card to sense the external environment.Therefore, the internal circuit state information contained in the original CSI data belongs to the CSI measurement error and needs to be eliminated.In this paper,the error source is analyzed in detail from the CSI amplitude and phase.The CSI amplitude error is corrected by windowing the data packet.The CSI phase error is corrected by using the sub-carrier phase difference between the two antennas.The corrected CSI data substantially eliminates the error introduced by the internal circuit state information,and provides real and effective data for the subsequent ranging scheme and fingerprint scheme.Furthermore,the existing ranging method based on the RSS/RSSI ranging schemes and the CSI obtained by independently using the three antennas has a low ranging accuracy and poor robustness.In this paper,the CSI information obtained on the 3 antennas is grouped in units of 2 to construct the matrix ? H[k],and the time domain sequence ? h[n] is proposed as a parameter to construct the ranging model.This time domain sequence characterizes the reverse convolution of the true CIR between the antennas.At the same time,Automatic Gain Control(AGC)will amplify the signal to different degrees according to the received signal strength.According to the experiment,the AGC will destroy the monotonic relationship between the ranging model and the distance.Therefore,this paper extracts the AGC information of the received data to eliminate it.The experimental results show that,compared with the ranging model constructed by using RSS/RSSI information as a parameter or independently using CSI information on 3 antennas as parameters,the localization accuracy and robustness of the proposed ranging model are improved.Finally,in the fingerprint scheme,this paper uses the sub-carrier CSI phase difference between the two antennas as the fingerprint feature.Compared with the use of CSI amplitude or phase on a single antenna as a fingerprint feature,the CSI phase difference stability and degree of freedom are higher,which can better characterize a specific position.In addition,unlike the traditional fingerprint scheme,this paper does not use fingerprint matching to determine the location of the target,but combines existing machine learning tools and uses Back Propagation(BP)neural network to train the fingerprint dataset to obtain a classification model.New fingerprints obtained during the online phase are provided to the trained classification model to predict the corresponding location.Compared with the fingerprint matching scheme,building a classification model can improve the real-time and robustness of the system.Integrated ranging and fingerprint scheme,this paper implements a single AP indoor localization system based on the combination of ranging and fingerprint on the commercial Intel 5300 network card.