Study On FM Fingerprint Positioning Technology

Fingerprint location technology is the key direction of indoor positioning research.The existing fingerprint location methods use the difference of signal strength RSS to create the corresponding relationship between RSS vector and position coordinates,and store the corresponding relationship in fingerprint database.Through the measured RSS vector of node,the optimization algorithm is used to estimate the corresponding position coordinates of the node.At present,indoor positioning signal sources mainly use Wi-Fi,Bluetooth,UWB and so on.These signal sources have disadvantages such as susceptibility to interference and high deployment and operation costs.On the contrary,FM broadcasting,as a kind of existing public wireless signal,has the characteristics of high power,frequency protection and wide area coverage,so it has unique advantages as a fingerprint location signal source.At present,the research on fingerprint location based on FM broadcasting has only seen some fragmented work,and no systematic research results.Based on the above requirements,this paper will carry out the research on fingerprint positioning system and its key technologies.Firstly,based on the characteristics of FM signal RSS,a more efficient fingerprint location model and algorithm are studied.Secondly,due to the change of indoor wireless communication environment and the nonlinear time-varying of FM signal itself,the fingerprint database data changes,so it is necessary to solve the problem of quickly updating the fingerprint database.Finally,in the face of the security requirements in the application process of fingerprint location,the location problems under various malicious attacks are studied.Based on the above three aspects,this paper studies the key technology of fingerprint location in FM broadcasting,and achieves the following achievements and innovative contents:Firstly,this paper analyzed the various signal characteristics(RSS,audio SNR,signal isolation,etc.)of FM broadcast through the measured data,and verify the signal strength RSS as a feature of fingerprint positioning.In view of the existing machine learning algorithms,this paper chooses KNN as the databasematching algorithm in the location fingerprint positioning stage.Because the fingerprint data of the adjacent reference points(RP)in the fingerprint database are similar,the fuzzy matching of the algorithm leads to the larger area surrounded by the selected RP points,which results in the low positioning accuracy when calculating these RP centroid coordinates.In view of the above problems,this paper proposes the KNN algorithm of weighted centroid.The RP point with the farthest distance from the centroid is replaced by the current RP centroid.Through multiple iterations,the positioning error is obviously reduced.Compared with the traditional WKNN,KNN and Bayesian positioning algorithms,the median positioning error is reduced by 32%,39% and 47% respectively.Secondly,due to many external factors(environmental factors,the change of the position of FM transmitting point,the time-varying of signal itself,etc.)that affect the accuracy of fingerprint data,a dynamic calibration algorithm of fingerprint database based on transfer learning model is proposed.By collecting a small amount of data of calibration reference points in different time periods,the real-time positioning of intelligent terminal is realized.This algorithm not only significantly improves the comprehensive performance of existing FM location algorithm,but also greatly saves the cost and time of manual survey fingerprint data.Thirdly,in view of the fact that nodes are vulnerable to malicious attacks in the process of dynamic sensor localization,this paper proposes a secure localization algorithm,which introduces FM security verification mechanism into the dynamic sensor Monte Carlo(MCL)algorithm,then accurately judges and deletes malicious anchor nodes by voting.The security algorithm can resist a variety of malicious attacks in the localization process of wireless sensor networks,which greatly reduces the positioning error of intelligent terminals under malicious attacks.