Research On Indoor Wlan Location Algorithm Based On SVM Classifier

Because the technology of WLAN is more and more mature and it has complementary with 3G, WLAN has become the high-speed wireless access method in indoor environment, especially in hotpots. In addition, with the popularity of wireless networks and rapid growth of intelligent mobile phones, location based services (LBS) have attracted more and more attention and shown great energy in many applications, such as emergency, medical care and customized information delivery. So the indoor locating technique based on WLAN has become global research hotpot. Among all kinds of indoor locating technology, indoor locating technology based on location fingerprints can be realized by the pure-software, meanwhile it has high positing precision and low cost, so it appears to offer clear advantages.This thesis makes a deep and systematic study on indoor locating technology based on location fingerprints, analyzed the restriction factor of improvement of posting precision, and propose two-step locating algorithm. First step RSS sample set is partitioned based on K-mean clustering method and SVM classifier. Second step precise localization is realized based on location fingerprints indoor locating algorithm. Two-step locating algorithm can reduce adverse affects of huge search space effective, and the validity of the proposed method is also proved in the thesis. The main works and innovations are as follows:At first, this thesis analyzed and compared three kinds of indoor locating algorithm based on location fingerprints which are widely used at present, including KNN (KNN , K Nearest Neighbors),neural network algorithm, SVR(Support Vector Regression).Then this thesis puts emphasis on analyzing the restriction factor of improvement of posting precision.Secondly, the principle of SVM two-class classifier and selecting method of SVM modal and parameters is study deeply in this thesis and improve algorithm which combined SVM two-class classifier with KNN classifier to prune RSS sample set was proposed against the weakness of SVM two-class classifier in actual application.At last, this thesis combined the result of RSS partition based on SVM classifier with KNN, neural algorithm, SVR to realize accurate localization. The result of simulation shows that the improved particle can effectively improve precision of location fingerprints algorithm and decrease the amount of computing.