The Enhancement Of Wi-Fi Fingerprint Positioning Using Deep Learning

With the popularity of Wireless Fidelity(Wi-Fi)and mobile intelligent terminals,people pay more and more attention to the locality and immediacy of information transmission.Under this demand,Location Based Service(LBS)emerged as the times require.The combination of LBS and mobile intelligent terminals has a great impact on people's consumption habits and lifestyles[1].According to the user's real-time positioning,the user is actively provided with the required service.This new service method is a location-based service.Now,LBS has played an important role in many areas such as road assisted navigation and medical services,as well as commercial advertising[2].Among them,location-based precision marketing has broad application prospects,and the realization of this goal requires precise positioning of users.The shop where you are.However,precise positioning techniques in indoor environments are still a challenging problem compared to indoor environments.Based on the research and analysis of the principle of sequence vectorization representation and deep learning model,this paper makes an in-depth study on the application of deep learning model to solve the problem of WLAN fingerprint location.The main research work of this paper is as follows:For the fingerprint preprocessing stage,there is a problem of large fingerprint noise and high fingerprint dimension.A fingerprint preprocessing method is proposed.The method includes AP selection and Wi-Fi representation:firstly,the AP is selected to remove unstable APs.Then,using the Wi-Fi2Vec mechanism,the Wi-Fi fingerprint is mapped into a low-dimensional vector,avoiding the high-latitude input causing the dimensional problem of the subsequent positioning model.At the same time,the vector trained by the Wi-Fi2Vec mechanism fully considers the context information,and as an input to the subsequent positioning model,the performance of the classifier is improved.In the positioning algorithm stage,the convolutional neural network structure suitable for Wi-Fi fingerprint sequence is proposed.The convolution kernel is used to extract the local features of the Wi-Fi fingerprint sequence,and the high-order features are extracted by the pooling layer,and then the extracted features are extracted.Higher-order features are used for Wi-Fi fingerprint positioning to improve the accuracy of Wi-Fi fingerprint positioning.In the fingerprint location stage,a attention mechanism model suitable for Wi-Fi fingerprint sequences is proposed.Firstly,the bidirectional cyclic neural network model is used to extract the Wi-Fi sequence,that is,the Wi-Fi data is used to express the Wi-Fi data.The attention model is used to determine the current input weight,that is,the current input is constructed.The similarity model with the target state,the greater the similarity between the two,the higher the weight of the current input.The weighted input is used for Wi-Fi fingerprint positioning to improve the accuracy of Wi-Fi fingerprint positioning.