Research On Location Technology Based On Cellular Network

With the development of mobile intelligent terminal and the popularization of wireless communication network, the research on the location of intelligent terminal has been paid more and more attention by learners at home and abroad. On the other hand, as we are living in the era of big data, the demand for location based services is increasing. A large number of users need navigation services and taxi travel services in the daily life, Based on historical location information, Server can provide users with personalized service and discovery potential users.Therefore, location information is significant, and a reliable and effective positioning technology is the key to achieve these needs.This paper mainly studies the localization performance of the fingerprint matching algorithm in the scene of the cellular network. The main contents of this paper include the establishment of reliable and effective fingerprint library, the different matching methods, the metric learning and the positioning model.First of all, we reviewed the commonly used positioning technology,the principle of each technology is introduced, and the advantages and disadvantages of each technology and the applicable scene are analysed according to their own technical characteristics. The methods of fingerprint matching and the density estimation which is commonly used in the probability matching model are introduced. Then, in order to improve the efficiency of fingerprint location technology, a method of establishing fingerprint database is proposed. Compared with the traditional k-nn algorithm, this algorithm has good performance. And we analysis the relationship between the maximum likelihood decision,maximum a posteriori decision and Nearest neighbor adjudication. By introducing the radial basis function and the LMNN algorithm in machine learning, the precision of positioning can be improved obviously. Then an improved probability model is proposed, and the joint probability of the base station combination is added into the objective function, which makes the model more robust and achieve a higher positioning accuracy.Finally, we analysed the performance of the model by using the laws of large numbers, central limit theorem and the Cramer-Rao bounds in statistics, and obtained the conclusion which is also consistent with our intuition. According to the Cramer-Rao bounds, the localization error of the model is transformed into the variance of the unbiased estimator, and obtained some conclusions about the accuracy of the model localization.