The Reseach Of Positioning Technology And System In Indoor Wi-Fi Environments

With the rapid development of indoor localization, localization technologies based on infrared ray, ultrasonic, Bluetooth, UWB(ultra wild band) and Wi-Fi(Wireless Fidelity), are becoming hot spots in academic and applied research. Indoor localization based on Wi-Fi has its own advantages in precision, robustness, security and complexity. This technology has received much concern since it can realize localization on any IT(intelligent terminal) in the form of APP(application), making use of the mature hardware platform and wireless network access points which can be found everywhere.This paper carried on research on indoor localization technology using Wi-Fi, including three parts: pre-processing of received signals, building position fingerprint database at the offline phase, and indoor localization at the online phase. Firstly, in order to get stable fingerprint data and raise the stability of the system, this paper used logarithm spectrum domain method for restraining multi-path effect to pre-process the received signal. The Kalman filtering method is also studied for pre-processing. Secondly, to reduce the high price for building fingerprint database, this paper applied matrix completion based on the SVT algorithm to reconstruct the fingerprint database with low rank. Clustering based on fast search of density peak is also proposed to refine the fingerprint database, which is compared with K-means and propagation clustering. Thirdly, this paper studied the localization algorithm using received signal strength filling based on the propagation model at online phase, with detailed research on the neighboring algorithm based on fingerprint matching, the Bayesian algorithm and the compressed sensing algorithm.Moreover, in this paper, an indoor localization system based on Wi-Fi is designed. In this system, pre-processing of received signal is based on logarithm spectrum domain restraining multi-path effect, clustering of fingerprint database is realized by fast search density peak clustering, and online localization using the Bayes algorithm. The experiments were implemented in real indoor environment with layout of reference points, which constructed the position fingerprint database of these experimental areas. The result obtained by applying the proposed system is 85% cumulative probability of localization error within two meters. This shows that the proposed localization system has good precision of localization and high robustness with less complexity.