Research Of Indoor Positioning Technology Based On Smart Terminals

As one of itie core technologies of Location Based Service (LBS),positioning technology has become a hotspot of current research in mobileIntenret area with the rapid popularization of smart terminals. Positioningdemand is also developed into more and more practical as from GPS-basedoutdoor to low-latency seamless indoor and outdoor positioning. In order toachieve and satisfy the positioning needs of various users and companies, novelmethods and means for positioning is constantly proposed and researched.Especially for the research of techniques of indoor positioning based on smartterminals. Since many traditional range-based positioning methods need foradditional hardware support and high deployment costs, range-free positioningmethods such as IP addresss positioning and fingerprints-based positioning aregetting more and more popular in practical use. In recent study, positioningmethod using WLAN receive signal strength (RSS) as fingerprints has becomethe most potential indoor positioning method for smart terminals because of itslow cost, wide coverage area and so on advantages. A lot of new methods hasbeen proposed basing on WLAN fingerprints positioning, many of which areconsidered from two sapects of innovation: one is the robusotess of thefingerpirnts database, namely establish a more effective fingerpirnts database forimproving the positioning performance of the system; the other one is the matching algorithm which extract a more precise positiong results with thelimited fingerpirnts database.Targeting on two aspect of innovation mentioned above, this paper mainlydoes research in the following fields based on indoor positioning technologiesfor smart terminals. First, considering on the unstableness of traditional RSSfingerpirnts under the indoor environment, we propose a novel design for thefingerpirnts by channel estimation method. Second, a more practical fingerprintsextraction technique based on kernel density estimation is proposed consideringthe diversity of smart terminals and experimented on a practical indoorpositioning platform which we designed for verifying. Third, we also propose anew positioning technique by pre-clustering the fingerprites database andcalculating the positioning result by cluster-matching and location-matching twostages of matching algorithm, which is an effective matching technique thatobviously reduces the response time and calculation complexity.How to reduce the cost and difficulty of the establishment and maintain thefingerpirnts database? The traditional method of indoor positioning systemrequires a large amount of work on collecting fingerprints. Thus, the firstresearch point of this paper is an indoor positioning system design based onWi-Fi fingerprints. More over, we propose two kinds of fingerprints collect andextract methods which include the channel estimation and kenrel densityestimation. Fingerpirnts collected and extracted with channel estimation methodsuccessfully reduce the unstableness of the RSS under indoor environmentwhich multi-channel effect inlfuence a log. Fingerprints collected and extractedwith kernel density estimation method mainly deat with the divercity of RSScollected by different kinds of smart terminals. Expeirments under those twomethods of collect and extract fingerpirnts have been achived in the designed indoor positioning system. Both simulation and experiment give that proposedfingerpritnts are more robust and stable than the traditional RSS fingerpritnts.Under a limited accurate fingerprint database, how to use those fingerpintsdata more effectively to calculate a more precise location result for the users in ashort time? The positioning algoirthm greatly affects the performance of thepositioning system in response time and computation complexity and futheraffects the users’ online experience. Existed works proposed a lot matchingalgorithms for calculating and searching the most similar sample location for the’users current location, including K-Nearest Neighbor (K-NN) algoirthm, fuzzylogic algorithm, pattern matching algorithm and so on. However, searching andmatching under a big data based fingerprints database is time consuming andcomputational costing, particularly when a lot users request for positioningconcurrently. Therefore, we propose a novel location matching method based onpre-clustered fingerpimts database which called K-NN combined with MMatching Clusters (MMC-KNN) algoirthm. This matching method effectivelyreduces the query operations compared with the traditional RSS matchingalgorithms, and further reduce itie positioning response time. From thesimulation and experimental results we can have that both the response time andpositioning accuracy significantly improved.