Research On WiFi Indoor Location Method Based On Location Fingerprint

In recent years,with the maturity of the fourth-generation network communication technology and the rapid development of the microelectronics industry,mobile terminal equipment has been widely popularized in people's daily lives.Therefore,based on the needs of Location Based Services(LBS)it is also increasing day by day,and at the same time,it places high requirements on positioning quality.Because the propagation of satellite signals is blocked by building shelters such as reinforced concrete in indoor environments,research on indoor positioning based on electromagnetic signals had begun to receive widespread attention.A loss model was established for the propagation of electromagnetic signals,which shows that Wireless Fidelity(Wi Fi)signals would have a certain quality attenuation in different areas during the propagation process.It is precisely by using the difference of this signal that the received signal strength(RSS)the positioning method of establishing electromagnetic fingerprint database had became one of the mainstream solutions.By reading and searching relevant domestic,foreign relevant materials and literature in recent years.Combining data mining technology and machine learning knowledge,the collected Wi Fi data was processed,modeled and analyzed.This study collects Wi Fi signal data through a dedicated electromagnetic signal acquisition tool,analyzes the attenuation characteristics of the Wi Fi signal,and then combines data mining knowledge and machine learning technology to propose an improved Wi Fi indoor fingerprint positioning method.The positioning process is divided into offline phase and online phase.This paper is in the offline phase of fingerprint location: First,it mainly cleans the collected Wi Fi data,including operations such as removing missing values and processing invalid data to further extract feature information.Second,the clustering method is mainly used to analyze the aggregation characteristics between the data to build a fingerprint database.In the online positioning stage,the current location information is predicted through the classifier in machine learning theory.The main innovations in the work described above are as follows:(1)This paper proposes an improved density peak cluster(K-IDPC)algorithm combined with KNN.The main content of the algorithm is as follows: In this paper,the ordinary density peak clustering algorithm is used to solve the problems of uneven density of positioning data and low discrimination between cluster centers.The correlation coefficient is used to indicate the degree of contribution of sample data to density.At the same time,the KNN idea is used to solve the problem of outliers in the clustering process,thereby improving the robustness to the positioning environment.In addition,this study combines the data segmentation algorithm to cut the offline electromagnetic signal data,and divides the large data set into several small data sets,which reduces the computational complexity of the improved clustering algorithm.(2)For the online phase,the traditional naive Bayes algorithm was usually used for location analysis.The algorithm assumes subjectivity and limitations on the independence assumption of the positioning data samples,and increases the problem of computational overhead.This paper proposes a posterior weighted na?ve bayes algorithm(PWNB).The main content of the algorithm is as follows: First,a Gaussian mixture model(GMM)is used to calculate the mutual influence of multiple Wi Fi overlay states at the positioning moment to obtain a weight coefficient to reflect the correlation between the positioning data.Secondly,by calculating the variance of the feature data in different positioning areas,and then determining whether the new positioning data needs to be calculated in the current round according to the fluctuation characteristics of the variance.This calculation process is called the posterior probability estimation process,which will significantly improve the immediacy of the classifier.Finally,based on the above theoretical methods and experimental analysis.According to the attenuation characteristics of the Wi Fi signal,the experimental results show that the traditional NB and DPC algorithm combined positioning accuracy is only 82%,and the proposed PWNB and K-IDPC algorithm combined positioning accuracy can reach 98%.And from the analysis of positioning time,using the algorithm proposed in this paper,the average positioning time is only 17.2s,which greatly improves the positioning efficiency.