Research On WLAN Indoor Location Alogrithm Based On Information Entropy

As a complement of broadband cable access network, WLAN is applied moreand more widely. That also spawned various services based on WLAN such asWLAN indoor positioning services. The WLAN indoor positioning system based onfingerprint becomes a hot research topic because of its simplicity and low cost. Thispaper studies WLAN indoor location methods based on fingerprint and through ourimproving methods we can improve the positioning accuracy and decrease thepositioning time.The WLAN indoor positioning technology based on fingerprint is divided intotwo phases: the offline phase which the radio map is created and online phasewhich the sample data is located. In the offline phase, a radio map is obtained bymeasuring in the real environment which stores the corresponding physical location.In the online phase the feature matching algorithms are used to obtain the onlinemeasured data’s physical location. The WLAN indoor positioning methods based onthe fingerprint need solve two problems: the positioning accuracy and the time usedto locate. Thus this paper proposed the clustering algorithms, AP selectionalgorithms and radio map updating algorithms.First, this paper analyzes the characteristics of the existing WLAN indoorpositioning systems based on fingerprint database. The methods of how to establishthe radio map and the matching algorithms are introduced. There are two measuresin the methods of establishing a radio map which are propagation model methodsand RSS feature value methods. And this paper selected the RSS feature valuemethods. The RSS value is changed over time, antenna orientation and thereference point changing. So we need use a reasonable method to establish a radiomap. In the feature matching algorithms the nearest neighbor algorithm, K-nearestneighbor algorithm and weighted K-nearest neighbor algorithm are introduced.Secondly, the paper analyzes the radio map. We put emphasis on studying howto simplify the radio map and how to update the radio map. To achieve the goal thatlocating timeliness, this paper first imported the clustering methods, after clustering,the radio map is divided into several categories. And then we use the AP selectionalgorithms to select the right combination of the APs to locate each subclass. In theclustering algorithms, we introduce the simplest K-means clustering algorithm, and considering the concept of membership the fuzzy K-means clustering algorithm isintroduced. These two algorithms need specify the number of cluster, so wepresented the affinity propagation clustering algorithm. In the AP selectionalgorithms, the random selection method and the mean maximum choices of APs,information entropy gain method and mutual information entropy are introduced.Finally, with the view of the accuracy of positioning, radio map updating methodsare introduced in which we use EM algorithm to solve the Hidden Markov model soas to update the radio map regular.Finally, we use feature matching algorithm to complete the positioning in ourreal environment. We analyzed the performance of the clustering algorithms, APselection algorithms and radio map updating algorithms. In the end, we select theappropriate parameters in our experimental environment thus achieve highpositioning accuracy and short positioning time costing.