Research On Indoor Location Method Based On Wi-Fi And Bluetooth Signal Source

With the rise of mature outdoor positioning and location-based service technology,positioning technology is widely used in various industries,which not only makes people’s lives more convenient,but also greatly enhances the commercial value of various industries.Although the global positioning system and Beidou satellite navigation can almost meet people’s needs for outdoor positioning,the above technologies cannot function in indoor environments due to the obstruction of building signals,which affects positioning accuracy.The existing indoor positioning technology has shortcomings such as low positioning accuracy and expensive positioning hardware requirements.In response to the above issues,this article takes indoor positioning in large-scale indoor environments as the background,and conducts research based on Wi-Fi and Bluetooth signals as positioning signal sources combined with deep learning technology.Firstly,in response to the problem of insufficient positioning accuracy in single mode positioning signal methods in large indoor environments,a Wi-Fi and Bluetooth fusion WB-Bayes positioning method based on convolutional neural networks and Bayesian estimation is proposed for indoor positioning.The WB-Bayes localization method is divided into two stages.In the initial localization stage,an automated architecture based on the VGG(Visual Geometry Group)module is used to optimize the convolutional neural network model by searching for the optimal value of hyperparameters.After completing the initial localization,the localization results are weighted and fused at the decision-making level using Bayesian estimation to complete the secondary localization.Secondly,to address the issue of weak adaptability of real-time localization algorithms for different signal sources in different environments,a deep learning model En-Hide-CNN based on convolutional neural networks is proposed for Wi-Fi/Bluetooth adaptive localization methods.This method uses a random forest regression to perform recursive feature elimination in the preprocessing stage,obtains general model parameters through pre training the model,and then applies the general model to other positioning tasks and fine tunes each task to improve performance to adapt to indoor positioning tasks under different signal source conditions.Finally,in order to verify the effectiveness of the two positioning methods proposed in this article in improving positioning accuracy,multiple sets of experiments were designed and implemented on multiple public datasets,and the experimental results were analyzed and summarized.