Research On Mobile Location Optimization Algorithm Based On Deep Learning

Location based services(LBS)attracts more and more attention with the increase of people's application methods.At the same time,the popularity of intelligent mobile terminal products puts forward higher requirements for LBS.LBS application is based on accurate location information technology,and an important part of accurate location information technology is real-time mobile positioning.At present,the most widely used is GPS(Global Positioning System)global positioning technology.Due to the complex environment of the city,the occlusion of buildings will make GPS invalid in the city road.In order to make up for the failure of GPS in urban road,this paper adopts the mobile positioning algorithm based on fingerprint information,uses the positioning technology of received signal strength indication(RSSI)in the spread signal and the fingerprint information of WIFI-AP in the base station to make up for the shortcomings of traditional positioning methods.However,the positioning accuracy will be greatly reduced.In order to better improve the success rate and accuracy of positioning,this paper uses deep learning method to deeply analyze the fingerprint information and the correlation between the points to be measured.As the clutter data has a great impact on the positioning accuracy,this paper proposes a data filtering method to eliminate the impact of clutter data on the positioning results.The data filtering algorithm is used to predict the location results by combining with the deep neural network and convolutional neural network,which can effectively improve the positioning accuracy and make up for the failure of GPS in urban roads.(1)This paper takes the North Campus of Fujian University of Technology as the experimental environment,collects RSSI data of base station and WIFI-AP in the experimental environment through mobile terminal equipment,and establishes the corresponding fingerprint database in the offline phase of fingerprint positioning algorithm.At the same time,a data filtering algorithm is proposed to filter the drift points and invalid WIFI-AP data.(2)The rise of deep learning adds more feasible solutions to problems in various fields.Neural network can deeply analyze the correlation between data and get some subtle characteristics.With this feature,this paper uses the deep neural network algorithm to analyze the association between RSSI data and location points,and compares it with the traditional fingerprint location matching algorithm,which can get more accurate positioning results.(3)Deep neural network has a good performance in analyzing the relevance of data.However,in the experiment,by deepening the number of network layers,the deep neural network did not achieve the expected more accurate experimental results.On the contrary,when the number of network layers reached a certain value,the positioning accuracy declined.This is because when the network is too deep,the input data can't affect the prediction results,that is,the gradient vanishing problem.In this paper,a cross layer connection method is proposed to solve the problem of gradient vanishing and further improve the positioning accuracy.Compared with deep neural network,convolution neural network has lower computational complexity,and can extract more features from fingerprint data in two-dimensional form,so as to improve the positioning accuracy.Therefore,this paper proposes a fingerprint positioning model based on convolution neural network,and solves the time-varying of data through the method of time series.