Improved K-nearest neighbor algorithm for indoor positioning using 5G channel state information

Abstract

As the demand for location-based services in indoor environments continues to grow, indoor positioning systems have attracted more and more attention. At present, the commonly used positioning system is based on the received signal strength of the WiFi network positioning algorithm. However, with the vigorous development of the fifth-generation(5G) mobile communication technology, the positioning accuracy can be greatly improved, and more accurate indoor positioning can be achieved. In order to improve the positioning accuracy, a positioning algorithm based on the channel state information reference signal is proposed. By using the smooth rank sequence (SRS) algorithm in the estimation of the number of received signal paths, the estimation accuracy of the angle of arrival(AoA) and the time of flight(ToF) can be improved. The combination of Gaussian Mixture Model(GMM) and the Expected Maximum(EM) algorithm further ensures the construction of offline fingerprint database. Besides, the online matching algorithm of fingerprint database is improved, and an improved k nearest neighbor algorithm is proposed. Compared with other matching algorithms, it proves that the proposed two-step k-nearest neighbor algorithm improves the positioning accuracy and reduces the amount of calculation for online matching.