Research On WiFi Indoor Positioning Based On DynFWA-SVM

With the rapid development of mobile Internet technologies and the widespread adoption of smart mobile terminal devices,people have given higher expectations and requirements for Location Based Services.At present,the positioning technology of the outdoor environment has matured,and the positioning technology of the indoor environment is still in the research stage.As the indoor positioning technology based on location fingerprint matching has the advantages of lower cost,simple hardware,and high positioning accuracy,it has become the research focus.There are urgent problems to be improved in the indoor positioning technology as follows:RSS fingerprints collected by different signal collection devices have obvious differences;WiFi signals are susceptible to interference in the process of indoor complex environment propagation;Large datasets require a lot of manpower cost when locating a fingerprint database;positioning algorithms support vector machine?SVM?with low positioning accuracy and time-consuming when the positioning area is large.To solve the above problems,this paper advances the plan as follows:?1?Using Gaussian filtering to filter the collected WiFi signal,and using SSD fingerprint to instead of the traditional RSS fingerprints.The theoretical analysis shows that the value of RSS is related to the antenna gain factorG_MT of the mobile terminal.Therefore,the signal strength of₁ collected at the same location and the signal strength of₂ are subtracted to eliminate thefactor,The experimental results show that the Gaussian filter-processed WiFi signal value is closer to its true value,and the difference of SSD fingerprints of different mobile devices is reduced by 83.1%compared with the difference of RSS fingerprints,which has better robustness.?2?Using Kriging interpolation method to establish large data amount location fingerprint database.High density fingerprint information can improve the positioning accuracy,but the indoor activity area is increasing.Meanwhile,the cost of building fingerprint database by human collection location fingerprint data is also increasing,therefore,in order to solve the above problems,the unknown points are interpolated by Kriging interpolation method,and the experimental results show that Kriging interpolation method can save 40%of the fingerprint information collection workload in offline phase.?3?With the increasing positioning area,the selection of penalty parameters and nuclear parameters of SVM will directly affect the positioning efficiency and positioning accuracy.Therefore,this paper introduces dynFWA to optimize its parameters,and uses SVM-based classification and regression positioning model to locate.Through contrast experiments,it is proved that dynFWA has the faster convergence rate and the better fitness function value in optimizing SVM parameters.The average positioning accuracy of dynFWA-SVM positioning model can reach 1.63 meters.