Font Size: a A A

Research Of Multi-antenna Localization Algorithm

Posted on:2018-04-02Degree:MasterType:Thesis
Country:ChinaCandidate:Y YuanFull Text:PDF
GTID:2348330512488250Subject:Engineering
Abstract/Summary:PDF Full Text Request
Wi-Fi based fingerprinting systems,mostly utilize the Received Signal Indicator(RSSI),which is known to be unreliable due to environmental and hardware effects.In this paper,we present a novel Wi-Fi fingerprinting system to realize a more accurate indoor localization,exploiting the fine-grained information known as Channel State Information(CSI)and kNN which is a method of machine learning.In this paper,the main research content includes the following.(1)We study the basic theory of indoor localization,introduce the existing indoor localization estimation and localization algorithm.Analyzing and Comparing the advantage of RSSI and CSI in Wi-Fi environment,and giving the reason why we choose CSI as the parametric of localization,especially emphasis on the wireless communication channel of 802.11 n.In the end,there is an improvement in CSI-based localization algorithm.(2)When choosing localization algorithm,we did not use the traditional method,such as three cutin method,hyperbolic method and least square method.But considering k-Nearest Neighbor and Bayes of the machine learning algorithm.On that basis,we simulate and analyze the localization performance based on CSI.,and adopt mean distance error and error of the cumulative distribution function to evaluate it.Results indicate that kNN has a more accuracy than Bayes.(3)On off-line phase,we establish the training fingerprinting which directly decide the localization performance when positioning.Hence it is crucial to deal with CSI and extract localization parameter after collecting CSI.In this paper,we propose a different algorithm to deal with CSI,and utilize PCA to reduce the dimension of CSI as a new parameter.Result shows that,the methods mentioned above both have the more excellent performance than traditional methods.Even more,the algorithm which used PCA get a higher accuracy.Utilizing simulation tool and experimental platform,we discuss different environment and the number of training samples how to affect the localization performance.In this paper,on the basic of theory analysis,we utilize Matlab and the CSI data collecting from the practical setting to make an investigation of the algorithms and parameters,and discuss some factors which effect the localization performance.In the end,the experimental result shows an accuracy improvement of 28% over CSI-MIMO with an accuracy of 0.863 meters.
Keywords/Search Tags:Wi-Fi indoor localization, CSI, kNN, PCA
PDF Full Text Request
Related items