Research On Indoor Position Fingerprint Location Method Based On DNN And Improved KNN Algorithm

With the rapid development of the internet and information technology,the advent of the era of mobile intelligence,in real life,people's travel can not be separated from positioning and navigation,car driving,travel route planning and other scenarios are closely related to positioning technology,of course,these are outdoor positioning.In addition,people's demand for indoor positioning-based services is also increasing,For example,large-scale supermarkets and merchants recommend advertisements based on the user's location,and provide real-time positioning of staff equipped with related equipment at the office site to facilitate the management of staff.How to effectively perform indoor positioning has important research value.WIFI provides convenience for people to quickly access the Internet due to its convenient deployment and low cost.At the same time,the rapid popularization of smart terminals makes it possible for people to use WIFI to achieve indoor positioning through smart hardware.In this thesis,aiming at the performance problems of indoor location fingerprint positioning,the optimization and improvement are performed from the offline and online stages respectively.First of all,in the offline stage,compared with traditional classification methods and designed a new DNN neural network model,the self-encoder network is used to construct and train the classifier network to ultimately reduce the traversal range of the pending fingerprint information and improve the accuracy of the prediction range.It lays the foundation for accurate positioning in the next stage.In the online positioning stage,it analyzes and compares the existing online positioning algorithms,fully considers the contribution of different fingerprints to the positioning,and introduces improved weighted KNN algorithm to give fingerprints different weights to assign different fingerprints to the fingerprint.It lays the foundation for accurate positioning in the next stage.In the online positioning stage,analyze and compare the existing online positioning algorithms,taking full account of the impact of different fingerprint contributions on positioning.By introducing a new weight calculation method to give different weights to the fingerprint,to avoid the use of fixed K value in the positioning stage Large errors and improve the accuracy of online positioning.Finally,experiments are performed on the UCI public data set,and the improved algorithm is compared with the mainstream indoor positioning algorithms.The experimental results prove that the improved algorithm can effectively improve the performance of indoor location fingerprinting.