Arama Görseller Haritalar Play YouTube Haberler Gmail Drive Daha fazlası »
Oturum açın
Ekran okuyucu kullanıcıları: Erişilebilirlik modu için bu bağlantıyı tıklayın. Erişilebilirlik modu aynı temel özelliklere sahiptir, ancak okuyucunuzla daha iyi çalışır.

Patentler

  1. Gelişmiş Patent Arama
Yayınlanma numarasıCN103648106 A
Yayın türüBaşvuru
Başvuru numarasıCN 201310750528
Yayın tarihi19 Mar 2014
Dosya kabul tarihi31 Ara 2013
Rüçhan tarihi31 Ara 2013
Şu şekilde de yayınlandı:CN103648106B
Yayınlanma numarası201310750528.6, CN 103648106 A, CN 103648106A, CN 201310750528, CN-A-103648106, CN103648106 A, CN103648106A, CN201310750528, CN201310750528.6
Buluş Sahipleri谭学治, 周才发, 马琳, 邓仲哲, 何晨光, 迟永钢, 魏守明
Başvuru sahibi哈尔滨工业大学
Alıntıyı Dışa AktarBiBTeX, EndNote, RefMan
Dış Bağlantılar:  SIPO, Espacenet
WiFi indoor positioning method of semi-supervised manifold learning based on category matching
CN 103648106 A
Özet
The invention discloses a WiFi indoor positioning method of semi-supervised manifold learning based on category matching, and relates to an indoor positioning method. The WiFi indoor positioning method disclosed by the invention is used for solving the problems that a Radio Map database is large and the like in an existing WiFi indoor positioning method. The WiFi indoor positioning method comprises the following steps: 1. collecting Radio Map; 2. carrying out intrinsic dimension analysis on the Radio Map; 3. carrying out clustering analysis on the Radio Map; 4. carrying out dimensionality reduction on the Radio Map; 5. adding RSS in the Radio Map to obtain Radio Mapul; and 6. carrying out dimensionality reduction on the Radio Mapul to obtain a characteristic transformation matrix V, and forming an online positioning database through the Radio Map * and V. The WiFi indoor positioning method also comprises the following steps: 1. online testing RSS; 2. carrying out dimensionality reduction on the RSS to obtain RSS *; 3. outputting a positioning result; and 4. updating the database. The WiFi indoor positioning method disclosed by the invention is applied to the field of network technology.
Hak Talepleri(6)  şu dilden çevrildi: Çince
1.一种基于类别匹配的半监督流形学习的WiFi室内定位方法,其特征在于基于类别匹配的半监督流形学习的WiFi室内定位方法离线阶段定位过程按以下步骤实现: 一、对待定位的室内区域布置AP,使无线信号覆盖待定位的室内区域,完成WiFi网络构建; 在待定位的室内区域规则选取并记录参考点的相应坐标,测量并依次记录参考点接收到的所有AP的RSS信号作为位置特征信息,构建Radio Map,并存储Radio Map ; 二、采用GMST本征维数估计算法对步骤一中构建的Radio Map的本征维数进行分析,得到的本征维数作为CM-SDE算法的输入参数之一,决定Radio Map降维后的维数; 三、采用KFCM算法对Radio Map进行聚类分析,实现建立的Radio Map的类别标记,并作为CM-SDE的输入参数之一,并且提供相应的初始聚类中心及类别标记; 四、步骤二中的本征维数与步骤三中的类别标记作为输入参数,采用CM-SDE算法对步骤一中构建的Radio Map降维,得出相应的降维后的RadioMap'RadioMap*作为在匹配定位数据库用于在线定位阶段; 五、将不同用户在线定位阶段测试得到的未标记RSS,采用类别匹配的方式加入至已有Radio Map中,得到相应的包含未标记信号覆盖图RadioMapul,通过类别匹配更新的聚类中心作为CM-SDE算法中新的类别输入参数; 六、步骤五中的更新的聚类中心作为输入参数,采用CM-SDE对RadioMapul降维得到特征变换矩阵V' , Y'与RadioMap*共同构成在线匹配定位数据库,用于在线阶段定位;其中,所述线阶段定位具体为: 六(一)、在线测试RSS ; 六(二)、采用V将RSS降维为RSS* ; 六(三)、采用KNN算法进行匹配定位输出定位结果; 六(四)、用户定位终端定位数据库更新; 即完成了一种基于类别匹配的半监督流形学习的WiFi室内定位方法的离线阶段实现方式。 A WiFi-based indoor positioning methods category matching semi-supervised manifold learning, wherein WiFi indoor positioning methods off stage of the positioning process flow based semi-supervised learning category matching shape to achieve the following steps: First, treat positioning Interior layout arrangement AP, so the wireless coverage area to be positioned interior, complete WiFi network construction; rule be located in the interior region of the select and record a corresponding reference point coordinates, measure and record turn reference point of all the received AP signal RSS feature information as the location to build Radio Map, and storage Radio Map; Second, the use GMST intrinsic dimension estimation intrinsic dimension algorithm Step one constructed Radio Map of analyze the intrinsic dimensionality was used as a CM-SDE Enter one of the parameters of the algorithm to determine the number of dimensions Radio Map dimensionality reduction; Third, the use of Radio Map KFCM algorithm clustering analysis to achieve the establishment of the category labeled Radio Map, and as one of the input parameters of CM-SDE, and provide appropriate initial cluster centers and category tags; four, step two of the intrinsic dimension and step three categories labeled as an input parameter, using the CM-SDE algorithm Step one constructed Radio Map dimensionality reduction, too the corresponding dimension reduction after RadioMap'RadioMap * as a matching location database for on-line positioning stage; five different users online positioning phase of the test was not marked RSS, matching methods use categories have been added to the Radio Map, to give the corresponding signal coverage maps contain unlabeled RadioMapul, by category matching the updated cluster centers as CM-SDE algorithm in new categories of input parameters; updated cluster centers six, step 5 as an input parameter, using the CM-SDE RadioMapul feature dimension reduction obtained on the transformation matrix V ', Y' and RadioMap * together form online matching location database for online stage location; wherein the wire positioning stage in particular: six (a), online testing RSS; six ( b) using V dimensionality reduction of the RSS RSS *; six (three), using KNN algorithm to match the positioning output positioning results; six (four), the user positioning terminal location database updates; complete a class-based matching half Offline phase supervised manifold learning WiFi indoor positioning method implementations.
2.根据权利要求1所述的一种基于类别匹配的半监督流形学习的WiFi室内定位方法,其特征在于步骤二中采用GMST本征维数估计算法对步骤一中构建的Radio Map的本征维数进行分析,其计算公式为: ^ntrinsicdim 测地距最小生成树算法1-α 上式中+中的a表示最小生成树的线性拟合表达式y=ax+b的斜率。 The WiFi indoor positioning method according to a class-based matching semi-supervised manifold learning claim, wherein step b used GMST intrinsic dimension estimation algorithm built in step one of this Radio Map intrinsic dimension for analysis, which is calculated as: ^ ntrinsicdim geodesic minimum spanning tree algorithm from the above formula 1-α + in a minimum spanning tree represents a linear fitting expression y = ax + b slope. 1-α 1-α
3.根据权利要求1所述的一种基于类别匹配的半监督流形学习的WiFi室内定位方法,其特征在于步骤四中生成新的降维后的信号覆盖图RadioMap*与步骤六中的V'表达式为: Radio Map*=V1.XX是需要降维的Radio Map。 3. WiFi indoor positioning method according to a class-based matching semi-supervised manifold learning claim, wherein the step coverage map generated four new dimension reduction after RadioMap * Step six of V 'expression is: Radio Map * = V1.XX is required dimensionality reduction Radio Map.
4.根据权利要求1所述的一种基于类别匹配的半监督流形学习的WiFi室内定位方法,其特征在于步骤五中类别匹配的实现,其核心流程分两步完成:第一步,寻找未标记RSS的类别属性,由下式完成类别属性标记: 4. The WiFi indoor positioning method according to a class-based matching semi-supervised manifold learning claim wherein the step of matching the realization fifth category, its core processes in two steps: first, search RSS unlabeled category property, by the following formula category attribute tags:
Figure CN103648106AC00031
第二步:对RSS进行门限检测:通过计算并判定广义符号值与门限值的关系,从而实现Radio Map及类别标记数据的更新,广义符号值及聚类中心的更新分别下两式完成: Step 2: the RSS threshold detection: By calculating and determining the generalized symbol of the relationship between the threshold values, enabling Radio Map and category marker data update, update generalized symbolic value and the cluster centers were completed under two formulas:
Figure CN103648106AC00032
其中,所述门限值VT=0.9N。 Wherein, the threshold value VT = 0.9N.
5.一种基于类别匹配的半监督流形学习的WiFi室内定位方法,其特征在于基于类别匹配的半监督流形学习的WiFi室内定位方法在线阶段定位过程通过下述步骤实现: 一、在线测试RSS ; 二、将得到的待定位点的RSS采用特征变换矩阵降维变换得到RSS* ; 三、采用KNN算法对RSS*与Radio Map*匹配位,对待定位点的具体位置坐标进行预测并进行在线数据的更新,其实现过程为: (1)在线阶段,测试点处接收的RSS= [AP1, AP2,…,APn],与特征变换矩阵V'相乘,从而得出降维后的RSS' =[AP1; AP2,…APd],其中d表示本征维数; (2)采用KNN算法实现RSS^与Radio Map*的匹配,采用与RSS^最近的K个参考点的坐标的平均值作为测试点,1'),其表达式为: A WiFi indoor positioning method based on semi-supervised learning flow category matching shape, characterized in that the WiFi indoor positioning methods online stage of the positioning process flow based semi-supervised learning category matching shape realized by the following steps: First, the online test RSS; Second, the resulting point is to be located using RSS feature dimension reduction transformation matrix transform RSS *; Third, the use of specific location coordinates KNN algorithm and Radio Map * RSS * Match bit, treat the anchor point to predict and online data updates, and its implementation process as follows: (1) Live stage, received at the test points RSS = [AP1, AP2, ..., APn], and feature transform matrix V 'multiplied to arrive after dimensionality reduction of RSS' = [AP1; AP2, ... APd], where d is the intrinsic dimension; (2) KNN algorithm RSS ^ and Radio Map * match, using the average of RSS ^ K coordinates of the nearest reference point as test point, 1 '), which was expressed as:
Figure CN103648106AC00033
式中,U',Y')为测试点预测的坐标,(Xi, Yi)为第i个近邻点的坐标,K为KNN算法中近邻的数目; 四、用户定位终端定位数据库更新,即完成了基于类别匹配的半监督流形学习的WiFi在线阶段室内定位方法。 Where, U ', Y') test point forecasts coordinates, (Xi, Yi) is the i-th coordinate neighbor points, K number of neighbors KNN algorithm; Fourth, the user positioning terminal location database update is complete the WiFi online stage indoor location based matching category semi-supervised manifold learning.
6.根据权利要求5所述的一种基于类别匹配的半监督流形学习的WiFi室内定位方法,其特征在于步骤三中对待定位点的具体位置坐标进行预测并进行在线数据的更新由权利要求I中所述的离线数据库和在线数据库实现: 离线定位数据库的实现方式为:将定位用户本次定位测得的RSS作为未标记数据采用类别匹配方式加入到Radio Map中,并在移动终端上实现对本地在线匹配定位数据库的更新,实现动态的更新本地的数据,从而实现离线数据库定位方式; 在线数据库的实现方式为:用户在线定位完成后,将用户本次在线测得的RSS值上传至在线定位数据库所在的服务器,并在服务器端将在线定位数据库进行更新将将在线定位数据传回上传RSS数据的定位终端。 6. WiFi indoor positioning method based on a five-category matching the semi-supervised manifold learning claim, wherein in step three specific location coordinates to predict treatment positioning point and line data are updated by the claim I in the offline database and online database implementation: implementation of the database is offline Location: The target users of this positioning measured using RSS as unlabeled data matching category was added to the Radio Map and implemented on the mobile terminal Live match updates on local positioning database, dynamic update local data, enabling offline database positioning mode; implementation of the online database is: users online positioning is completed, the user of this line measured RSS values uploaded to online the server where the database is located, and upload the positioning terminal RSS data on the server side online location database updates will be online positioning data back.
Açıklama  şu dilden çevrildi: Çince

—种基于类别匹配的半监督流形学习的WiFi室内定位方法 - Kind of WiFi indoor positioning method based on semi-supervised stream category matching shape learning

技术领域 Technical Field

[0001] 本发明涉及一种室内定位方法,具体涉及一种基于类别匹配的半监督流形学习的WiFi室内定位方法。 [0001] The present invention relates to a method of indoor positioning, particularly to WiFi indoor positioning category matching method based on semi-supervised manifold learning.

背景技术 Background

[0002] 随着无线局域网络在世界范围的飞速发展和移动终端设备的广泛普及,近年来出现了许多室内定位相关的技术和应用。 [0002] With the widespread popularity of wireless local area networks in the world and the rapid development of mobile terminal equipment, in recent years there has been a lot of indoor positioning technology and related applications. 由于多径效应、信号衰减及室内定位环境的复杂性,基于传统的信号传播模型的室内定位方法难以达到高精度的室内定位要求。 Due to multipath effects, attenuation and complexity of indoor positioning environment, it is difficult to achieve high accuracy indoor location-based indoor positioning method requires conventional signal propagation model. 基于到达时间(Time of Arrival),到达时间差(Time Difference of Arrival)和到达角度(Angles ofArrival)等定位方法虽然可以基本满足定位精度需求,然而都需要定位终端有额外的硬件设备支持,具有较大局限性,从而导致基于上述几类定位方法的室内定位系统没有得到普及。 Based on the time of arrival (Time of Arrival), arrival time difference (Time Difference of Arrival) and angle of arrival (Angles ofArrival) method, although other location positioning accuracy can basically meet the demand, however, we will need to locate the terminal has additional hardware support, with greater limitations, resulting in not been universal indoor localization system based on the above-mentioned types of positioning method.

[0003] 目前,基于WLAN位置指纹(Finger Print)的WiFi室内定位方法得到了广泛应用。 [0003] Currently, WLAN based on the location of the fingerprint (Finger Print) of WiFi indoor positioning method has been widely used. 该方法的网络构建方法成本低廉,其使用2.4GHz ISMdndustrial Science Medicine)公共频段且无需在现有设施之上添加定位测量专用硬件。 The method of construction of a network of low cost, using 2.4GHz ISMdndustrial Science Medicine) public band and positioning measurement without adding dedicated hardware on existing facilities. 只需要通过移动终端的无线网卡及相应软件测量接收到的接入点(Access Point, AP)的信号强度(Received SignalStrength, RSS),由此来构建网络信号覆盖图(Radio Map),进而通过匹配算法来预测移动用户所处位置的坐标,或相对位置。 Only received through the wireless network card of the mobile terminal and an access point corresponding to the measurement software (Access Point, AP) signal strength (Received SignalStrength, RSS), to thereby construct the network signal coverage map (Radio Map), and further by matching algorithm to predict the coordinates of the mobile user's location, or relative position.

[0004] 然而通过该方式建立的Radio Map包含有庞大的数据信息,且随着定位区域扩大,Radio Map可能(依据定位匹配方式及算法选择)呈指数形势增长。 [0004] However, the manner established by Radio Map contains a large data information, and with the expansion of the location area, Radio Map may (by targeting and matching algorithm selection) exponential growth situation. 获得尽可能多的相关数据特征信息对于整个系统来说会提升定位精度,但是处理大量的特征信息增加算法开销,定位算法无法在处理能力有限的移动终端上有效运行,同时某些特征信息可能是对于定位没有作用甚至有负面作用,致使匹配效率降低,从而导致匹配定位算法的实现变得更加复杂,并且定位精度下降。 Get as much relevant data attribute information for the entire system, will increase the positioning accuracy, but a large number of characteristic information processing algorithms to increase spending, positioning algorithm can not operate effectively on a limited processing capability of the mobile terminal, while some feature information may be For positioning without effect even have a negative effect, resulting in reduced matching efficiency, thus leading to the matching location algorithm becomes more complicated, and positioning accuracy.

[0005] 当AP的数目增加及定位的参考点(Reference Point)增加时,Radio Map的数据信息增加。 [0005] When the number of AP and the positioning of the reference point (Reference Point) increase, increase Radio Map data information. 此时,Radio Map中代表的AP数目的信息表示了数据的维数。 In this case, AP Radio Map information on the number of representatives expressed the dimension of the data. 因此,当AP数目增加,Radio Map就变成了高维数据。 Thus, when increasing the number of AP, Radio Map it becomes a high-dimensional data. 为减轻处理高维数据的负担,降维算法是有效的解决方法之一。 To alleviate the burden of high-dimensional data dimensionality reduction algorithm is an effective solution. 高维数据可能包含很多特征,这些特征都在描述同一个事物,这些特征一定程度上是紧密相连的。 High-dimensional data may contain many characteristics that are describing the same thing, to some extent, these characteristics are closely linked. 如当从各个角度对同一个物体同时拍照时,得到的数据就含有重叠的信息。 As when viewed from different angles on the same object at the same time taking pictures, the data obtained to contain overlapping information. 如果能得到这些数据的一些简化的不重叠的表达,将会极大地提高数据处理运行的效率并一定程度上提高准确度。 If you can get some simplified expression does not overlap the data, will greatly improve the efficiency of data processing operation and improve the accuracy of a certain degree. 降维算法的目的也正是在于提高高维数据的处理效率。 Dimensionality reduction algorithm is exactly the purpose is to improve the processing efficiency of high-dimensional data.

[0006] 除了可以简化数据使其能够高效处理外,降维方法还可以实现数据可视化。 [0006] In addition to simplifying the data so that it can efficiently handle external dimension reduction method may also be data visualization. 由于很多统计学的和机器学习算法对于最优解的准确性很差,降维的可视化应用可以令用户能够实际看到高维数据的空间结构和算法输出的能力,具有很强的应用价值。 Since many statistical and machine learning algorithms for optimal accuracy is poor, lower-dimensional visualization applications can make the user can actually see the output capacity of the spatial structure and algorithm of high dimensional data, it has a strong value.

[0007]目前有很多基于不同目的的降维算法,包括有线性与非线性降维算法。 [0007] There are a lot of dimension reduction algorithm based on different purposes, including linear and nonlinear dimensionality reduction algorithm. 其中PCA(Principal Component Analysis)和LDA(Linear Discriminant Analysis)是典型的线性降维算法。 Wherein the PCA (Principal Component Analysis) and LDA (Linear Discriminant Analysis) is a typical linear dimension reduction algorithm. 这一类算法对于具有线性结构的高维数据有着良好的处理结果,但对于非线性结构的高维数据没有好的结果。 This class of algorithm for high-dimensional data with a linear structure with good results, but for non-linear high-dimensional data structure is not a good result. 典型的非线性降维算法以流形学习(ManifoldLearning)算法。 A typical nonlinear dimensionality reduction algorithm manifold learning (ManifoldLearning) algorithm. 2000年Science杂志上同一期发表了3篇有关于流形学习算法中提出了2 种经典的流形学习算法:LLE (Local Linear Embedding)及ISOMAP (Isometric Mapping)。 2000 on the same issue of Science magazine published three on manifold learning algorithms have been proposed two kinds of classic manifold learning algorithms: LLE (Local Linear Embedding) and ISOMAP (Isometric Mapping). 由此,各种基于不同的准则的流形学习算法被提出并有一部分流形学习算法应用于图像处理方面。 Thus, a variety of manifold learning algorithm based on different criteria have been proposed and partially manifold learning algorithms for image processing. LDE (Local Discriminant Embedding)算法是流形学习算法中较晚提出的,它是一种的典型的基于特征提取的流形学习算法,而不只基于可视化目标。 LDE (Local Discriminant Embedding) algorithm is manifold learning algorithm is proposed later, it is a typical manifold learning algorithms based feature extraction, not only based on visual target.

[0008] 对于上述的降维算法并不能对在线获得的RSS数据或者新增的RSS来提高降维的精度。 [0008] For the above dimension reduction algorithm and can not get RSS data online or new RSS to improve the accuracy of dimension reduction. 由于室内定位环境的变化,在不同时刻,特别是在长时间后,RSS数据之间的相关性就会降低。 Due to changes in indoor positioning environment, at different times, especially after a long correlation between the RSS data is reduced. 在已有算法中,有一类算法可以将在不同时间的RSS数据同时加入相应的已有数据中,从而增强不同数据之间的相关性,提高降维精度。 In the existing algorithm, there is a class of algorithms can be added to the corresponding existing data simultaneously at different times of RSS data, thereby enhancing the correlation between different data to improve the accuracy of dimension reduction. 这一类算法通常称之半监督算法。 This class algorithms typically termed semi-supervised algorithm. 根据半监督算法的特点,提出半监督鉴别嵌入(Sem1-supervised DiscriminantEmbedding, SDE )算法。 According to the characteristics of semi-supervised algorithm, semi-supervised identification embedded (Sem1-supervised DiscriminantEmbedding, SDE) algorithm.

[0009] 根据SDE算法的特点,采用类别匹配的方式将在线新得到的RSS加入已有RadioMap,然后进行局部鉴别嵌入降维,从而提出基于类别匹配的半监督局部鉴别嵌入算法(Classification Matching Based Sem1-supervised Discriminant Embedding,CM-SDE)。 [0009] According to the characteristics SDE algorithm, using categories that match the way the new line was added to the existing RSS RadioMap, then partial identification of embedding dimensionality reduction, which made embedded algorithm (Classification Matching Based Sem1 categories based on semi-supervised local matching identification -supervised Discriminant Embedding, CM-SDE). 采用CM-SDE算法对Radio Map进行降维,得出降维后的Radio Map,将对降维后的RadioMap室内定位,从而提出基于CM-SDE算法的WiFi室内定位算法。 Using the CM-SDE Radio Map algorithm to reduce the dimension after dimension reduction obtained Radio Map, will drop dimensionality RadioMap indoor location, which made WiFi indoor positioning algorithm based CM-SDE algorithms.

发明内容 DISCLOSURE

[0010] 本发明是要解决现有WiFi室内定位方法中存在的Radio Map数据库大,以及由于在线定位阶段计算复杂度高而引起的难以应用在线阶段获得RSS数据、难于在移动终端实现以及难于保证定位的实时性要求等问题,而提供了一种基于类别匹配的半监督流形学习的WiFi室内定位方法。 [0010] The present invention is to solve the existing WiFi indoor positioning methods exist Radio Map database is large, and difficult to apply online because the online positioning stage phase high computational complexity caused by the RSS data obtained, it is difficult to achieve in the mobile terminal and difficult to ensure Location of real-time requirements and other issues, and provides a WiFi indoor positioning category matching method based on semi-supervised manifold learning.

[0011] 基于类别匹配的半监督流形学习的WiFi室内定位方法离线阶段定位过程按以下步骤实现: [0011] WiFi indoor positioning methods off stage of the positioning process is based on the category matching the semi-supervised manifold learning to achieve the following steps:

[0012] 一、对待定位的室内区域布置AP,使无线信号覆盖待定位的室内区域,完成WiFi网络构建; [0012] First, treat positioning indoor areas arranged AP, so the wireless coverage area to be located in the interior, complete WiFi network construction;

[0013] 在待定位的室内区域规则选取并记录参考点的相应坐标,测量并依次记录参考点接收到的所有AP的RSS信号作为位置特征信息,构建Radio Map,并存储Radio Map ; [0013] The rule is to be positioned in the interior region of the select and record a corresponding reference point coordinates, measure and record turn reference point of all the received AP RSS signal characteristic information as the location to build Radio Map, and storage Radio Map;

[0014] 二、采用GMST本征维数估计算法对步骤一中构建的Radio Map的本征维数进行分析,得到的本征维数作为CM-SDE算法的输入参数之一,决定Radio Map降维后的维数; [0014] Second, the use GMST intrinsic dimension intrinsic dimension algorithm to step a built in Radio Map of estimation analysis, one intrinsic dimension as CM-SDE algorithm input parameters obtained decide Radio Map drop dimension after dimension;

[0015] 三、采用KFCM算法对Radio Map进行聚类分析,实现建立的Radio Map的类别标记,并作为CM-SDE的输入参数之一,并且提供相应的初始聚类中心及类别标记; [0015] Third, the use of Radio Map KFCM algorithm clustering analysis to achieve the establishment of the category labeled Radio Map, and as one of the input parameters CM-SDE, and provide appropriate initial cluster centers and category tags;

[0016] 四、步骤二中的本征维数与步骤三中的类别标记作为输入参数,采用CM-SDE算法对步骤一中构建的Radio Map降维,得出相应的降维后的RadioMap'RadioMap*作为在匹配定位数据库用于在线定位阶段; [0016] Fourth, step two and step intrinsic dimension of the three categories of mark as an input parameter, using the CM-SDE algorithms built in one step Radio Map dimensionality reduction, draw the corresponding RadioMap after dimensionality reduction ' RadioMap * as a matching location database for on-line positioning stage;

[0017] 五、将不同用户在线定位阶段测试得到的未标记RSS,采用类别匹配的方式加入至已有Radio Map中,得到相应的包含未标记信号覆盖图RadioMapul,通过类别匹配更新的聚类中心作为CM-SDE算法中新的类别输入参数; [0017] Five different users online positioning phase of the test was not marked RSS, matching methods use category added to the existing Radio Map obtain the corresponding cluster center contains unlabeled coverage map RadioMapul, by category match updates CM-SDE algorithms as new categories of input parameters;

[0018] 六、步骤五中的更新的聚类中心作为输入参数,采用CM-SDE对RadioMapul降维得到特征变换矩阵V' ,Ψ与RadioMap*共同构成在线匹配定位数据库,用于在线阶段定位;其中,所述线阶段定位具体为: [0018] VI, the fifth step of updating the cluster centers as an input parameter, using the CM-SDE to get RadioMapul feature dimension reduction transformation matrix V ', Ψ and RadioMap * together form online matching location database for online stage positioning; wherein the wire positioning stage in particular:

[0019] 六(一)、在线测试RSS ; [0019] VI (a), online test RSS;

[0020] 六(二)、采用V将RSS降维为RSS* ; [0020] VI (b), the use of V dimension reduction of the RSS RSS *;

[0021] 六(三)、采用KNN算法进行匹配定位输出定位结果; [0021] VI (III), using KNN algorithm to match the positioning output positioning results;

[0022] 六(四)、用户定位终端定位数据库更新; [0022] VI (D), the user positioning terminal location database updates;

[0023] 即完成了一种基于类别匹配的半监督流形学习的WiFi室内定位方法的离线阶段实现方式。 [0023] to complete the offline phase matching categories based semi-supervised manifold learning WiFi indoor positioning method of implementation.

[0024] 基于类别匹配的半监督流形学习的WiFi室内定位方法在线阶段定位过程通过下述步骤实现: [0024] WiFi indoor positioning methods online stage of the positioning process is based on the category of semi-supervised matching manifold learning through the following steps to achieve:

[0025] 一、在线测试RSS; [0025] First, the online test RSS;

[0026] 二、将得到的待定位点的RSS采用特征变换矩阵降维变换得到RSS* ; [0026] Second, the resulting point is to be located using RSS feature dimension reduction transformation matrix transform RSS *;

[0027] 三、采用KNN算法对RSS*与Radio Map*匹配位,对待定位点的具体位置坐标进行预测并进行在线数据的更新,其实现过程为: [0027] Third, the use of specific location coordinates KNN algorithm and Radio Map * RSS * Match bit, treat the anchor point to predict and online data update, and its implementation process as follows:

[0028] (I)在线阶段,测试点处接收的RSS= [AP1, AP2,…,APn],与特征变换矩阵V'相乘,从而得出降维后的RSS' =[AP1; AP2,…APd],其中d表示本征维数; [0028] (I) Live stage, received at the test points RSS = [AP1, AP2, ..., APn], and feature transform matrix V 'multiplied to arrive after dimensionality reduction of RSS' = [AP1; AP2, ... APd], where d is the intrinsic dimension;

[0029] (2)采用KNN算法实现RSS^与Radio Map*的匹配,采用与RSS^最近的K个参考点的坐标的平均值作为测试点(X' ,Y'),其表达式为: [0029] (2) The KNN algorithm RSS ^ and Radio Map * match, using the RSS ^ mean coordinates K nearest reference point as a test point (X ', Y'), which was expressed as:

[0030] [0030]

Figure CN103648106AD00061

[0031] 式中,(X' , y')为测试点预测的坐标,(Xi, y)为第i个近邻点的坐标,K为KNN算法中近邻的数目; [0031] wherein, (X ', y') coordinates of the test point forecast, (Xi, y) coordinates of the i-th neighbor points, K is the number of neighbors KNN algorithm;

[0032] 四、用户定位终端定位数据库更新,即完成了基于类别匹配的半监督流形学习的WiFi在线阶段室内定位方法。 [0032] Fourth, the user positioning terminal location database update, which completed the WiFi online stage indoor location based on semi-supervised stream category matching shape learning.

[0033] 发明效果: [0033] Effect of the Invention:

[0034] 针对本发明提出的CM-SDE算法及背景需要求,通过对哈尔滨工业大学科学园2A栋12层走廊构成的室内定位区域进行定位。 [0034] For CM-SDE algorithms and background of the present invention provides a request, through the indoor positioning region Harbin Institute of Technology Science Park Building, 12th floor corridor 2A constituted positioning. 采用联想V450笔记本电脑并结合NetStumbler软件测试所有参考点处的RSS,构成Radio Map,并采用CM-SDE算法对Radio Map降维并采用KNN算法实现室内定位。 Lenovo V450 notebook computers using NetStumbler software testing in conjunction with all of the reference point RSS, constitute Radio Map, and the use of CM-SDE Radio Map dimensionality reduction algorithm and KNN algorithm using indoor positioning. 在附图4的仿真中,降维后的Radio Map的维数为原始RadioMap维数的三分之一。 In the drawings, the simulation 4, reducing the dimension of one third of the original RadioMap dimension is the dimension of post-Radio Map. 从附图4的仿真结果来看,采用CM-SDE算法与初始的KNN的算法的定位性能可比拟,但CM-SDE的定位复杂度仅为三分之一,且CM-SDE算法可以有效地应用实时得到的新的RSS来Radio Map的密度,从而有效的提高定位精度。 Simulation results from the figures 4, the positioning performance using CM-SDE algorithms and initial KNN algorithm can be compared, but the positioning of the complexity of the CM-SDE is only one-third, and the CM-SDE algorithm can effectively real-time to get the new RSS to Radio Map of density, which effectively improve positioning accuracy.

附图说明 Brief Description

[0035] 图1是本发明中离线数据库定位实施流程图;实线箭头表示步骤之间的数据传输;[0036] 图2是本发明中在线数据库定位实施流程图;实线箭头表示步骤之间的数据传输; [0035] FIG. 1 is positioned in the present invention flowchart showing an offline database; solid line arrows indicate data transmission between steps; [0036] FIG. 2 is an online database of the present invention, a flow chart for positioning; the solid arrow represents a step between data transfer;

[0037] 图3是基于WiFi的室内定位网络的构建及实验环境示意图; [0037] FIG. 3 is a schematic diagram of the experimental environment to build and WiFi network based on indoor positioning;

[0038] 图4是采样网格图; [0038] FIG. 4 is a sampling grid;

[0039] 图5是采用CM-SDE与KNN算法定位性能对比图;其中,表示CM-SDE,表示 [0039] FIG. 5 is a CM-SDE KNN algorithm and positioning performance comparison chart; wherein, represents CM-SDE, represents

KNN。 KNN.

具体实施方式 DETAILED DESCRIPTION

[0040] 具体实施方式一:本实施方式的基于类别匹配的半监督流形学习的WiFi室内定位方法离线阶段定位过程按以下步骤实现: [0040] A specific embodiment: This embodiment WiFi indoor positioning methods off stage of the positioning process is based on the category of semi-supervised matching manifold learning to achieve the following steps:

[0041] 一、对待定位的室内区域布置AP,使无线信号覆盖待定位的室内区域,完成WiFi网络构建; [0041] First, treat positioning indoor areas arranged AP, so the wireless coverage area to be located in the interior, complete WiFi network construction;

[0042] 在待定位的室内区域规则选取并记录参考点的相应坐标,测量并依次记录参考点接收到的所有AP的RSS信号作为位置特征信息,构建Radio Map,并存储Radio Map ; [0042] The rule is to be positioned in the interior region of the select and record a corresponding reference point coordinates, measure and record turn reference point of all the received AP RSS signal characteristic information as the location to build Radio Map, and storage Radio Map;

[0043] 二、采用GMST本征维数估计算法对步骤一中构建的Radio Map的本征维数进行分析,得到的本征维数作为CM-SDE算法的输入参数之一,决定Radio Map降维后的维数; [0043] Second, the use GMST intrinsic dimension intrinsic dimension algorithm to step a built in Radio Map of estimation analysis, one intrinsic dimension as CM-SDE algorithm input parameters obtained decide Radio Map drop dimension after dimension;

[0044] 三、采用KFCM算法对Radio Map进行聚类分析,实现建立的Radio Map的类别标记,并作为CM-SDE的输入参数之一,并且提供相应的初始聚类中心及类别标记; [0044] Third, the use of Radio Map KFCM algorithm clustering analysis to achieve the establishment of the category labeled Radio Map, and as one of the input parameters CM-SDE, and provide appropriate initial cluster centers and category tags;

[0045] 四、步骤二中的本征维数与步骤三中的类别标记作为输入参数,采用CM-SDE算法对步骤一中构建的Radio Map降维,得出相应的降维后的RadioMap'RadioMap*作为在匹配定位数据库用于在线定位阶段; [0045] Fourth, step two and step intrinsic dimension of the three categories of mark as an input parameter, using the CM-SDE algorithms built in one step Radio Map dimensionality reduction, draw the corresponding RadioMap after dimensionality reduction ' RadioMap * as a matching location database for on-line positioning stage;

[0046] 五、将不同用户在线定位阶段测试得到的未标记RSS,采用类别匹配的方式加入至已有Radio Map中,得到相应的包含未标记信号覆盖图RadioMapul,通过类别匹配更新的聚类中心作为CM-SDE算法中新的类别输入参数; [0046] Five different users online positioning phase of the test was not marked RSS, matching methods use category added to the existing Radio Map obtain the corresponding cluster center contains unlabeled coverage map RadioMapul, by category match updates CM-SDE algorithms as new categories of input parameters;

[0047] 六、步骤五中的更新的聚类中心作为输入参数,采用CM-SDE对RadioMapul降维得到特征变换矩阵V' ,Ψ与RadioMap*共同构成在线匹配定位数据库,用于在线阶段定位;其中,所述线阶段定位具体为: [0047] VI, the fifth step of updating the cluster centers as an input parameter, using the CM-SDE to get RadioMapul feature dimension reduction transformation matrix V ', Ψ and RadioMap * together form online matching location database for online stage positioning; wherein the wire positioning stage in particular:

[0048] 六(一)、在线测试RSS ; [0048] VI (a), online test RSS;

[0049] 六(二)、采用V将RSS降维为RSS* ; [0049] VI (b), the use of V dimension reduction of the RSS RSS *;

[0050] 六(三)、采用KNN算法进行匹配定位输出定位结果; [0050] VI (III), using KNN algorithm to match the positioning output positioning results;

[0051] 六(四)、用户定位终端定位数据库更新; [0051] VI (D), the user positioning terminal location database updates;

[0052] 即完成了一种基于类别匹配的半监督流形学习的WiFi室内定位方法的离线阶段实现方式。 [0052] to complete the offline phase matching categories based semi-supervised manifold learning WiFi indoor positioning method of implementation.

[0053] 所述离线阶段与在线阶段均在定位终端完成。 [0053] The off-line and on-line stage stages are completed in the positioning terminal.

[0054] 具体实施方式二:本实施方式与具体实施方式一不同的是:步骤二中采用GMST本征维数估计算法对步骤一中构建的Radio Map的本征维数进行分析,其计算公式为: [0054] DETAILED DESCRIPTION II: The embodiment of a specific embodiment is different: Step two adopted GMST intrinsic dimension intrinsic dimension estimation algorithm Step One Radio Map constructed to analyze, the calculation formula as follows:

[0055] d dim =+,测地距最小生成树算法 [0055] d dim = +, geodesic distance minimum spanning tree algorithm

1-α[0056] 上式中1 中的a表示最小生成树的线性拟合表达式y=ax+b的斜率。 1-α [0056] In the above formula 1 in a minimum spanning tree represents a linear fitting expression y = ax + b slope.

1-α 1-α

[0057] 其它步骤及参数与具体实施方式一相同。 [0057] Other steps and parameters with a specific embodiment the same.

[0058] 具体实施方式三:本实施方式与具体实施方式一或二不同的是:步骤四中生成新的降维后的信号覆盖图RadioMap*与步骤六中的V'表达式为: [0058] DETAILED DESCRIPTION three: the present embodiment and the detailed description or the second one is different: step four to generate a new low dimensional signal coverage map RadioMap * Step six of the V 'expression:

[0059] Radio Map*=V/.X [0059] Radio Map * = V / .X

[0060] X是需要降维的Radio Map。 [0060] X is a need to reduce the dimension Radio Map.

[0061 ] 其它步骤及参数与具体实施方式一或二相同。 [0061] the same one or two steps and other parameters of particular embodiments.

[0062] 具体实施方式四:本实施方式与具体实施方式一至三之一不同的是:步骤五中类别匹配的实现,其核心流程分两步完成: [0062] DETAILED DESCRIPTION Four: This embodiment is one of one to three particular embodiment is different: Fifth category matching step realization of its core processes in two steps:

[0063] 第一步,寻找未标记RSS的类别属性,由下式完成类别属性标记: [0063] First, look for unlabeled RSS categories attribute mark by the formula category properties:

Figure CN103648106AD00081

[0065] 第二步:对RSS进行门限检测:通过计算并判定广义符号值与门限值的关系,从而实现Radio Map及类别标记数据的更新,广义符号值及聚类中心的更新分别下两式完成: [0065] Step 2: RSS perform threshold detection: By calculating and determining the relationship between generalized symbolic value with a threshold value, thereby achieving Radio Map and category marker data update, update generalized symbolic value and were under two cluster centers type finish:

Figure CN103648106AD00082

[0068] 其中,所述门限值VT=0.9Ν。 [0068] wherein, the threshold value VT = 0.9Ν.

[0069] 其它步骤及参数与具体实施方式一至三之一相同。 [0069] One same one to three other steps and parameters and specific embodiments.

[0070] 具体实施方式五:基于类别匹配的半监督流形学习的WiFi在线阶段室内定位方法通过下述步骤实现: [0070] DETAILED DESCRIPTION five: WiFi Live Stage indoor location based matching category semi-supervised manifold learning realized by the following steps:

[0071] 一、在线测试RSS; [0071] First, the online test RSS;

[0072] 二、将得到的待定位点的RSS采用特征变换矩阵降维变换得到RSS* ; [0072] Second, the resulting point is to be located using RSS feature dimension reduction transformation matrix transform RSS *;

[0073] 三、采用KNN算法对RSS*与Radio Map*匹配位,对待定位点的具体位置坐标进行预测并进行在线数据的更新,其实现过程为: [0073] Third, the use of specific location coordinates KNN algorithm and Radio Map * RSS * Match bit, treat the anchor point to predict and online data update, and its implementation process as follows:

[0074] (I)在线阶段,测试点处接收的RSS= [AP1, AP2,…,APn],与特征变换矩阵V'相乘,从而得出降维后的RSS' =[AP1; AP2,…APd],其中d表示本征维数; [0074] (I) Live stage, received at the test points RSS = [AP1, AP2, ..., APn], and feature transform matrix V 'multiplied to arrive after dimensionality reduction of RSS' = [AP1; AP2, ... APd], where d is the intrinsic dimension;

[0075] (2)采用KNN算法实现RSS^与Radio Map*的匹配,采用与RSS^最近的K个参考点的坐标的平均值作为测试点(X' ,1'),其表达式为: [0075] (2) The KNN algorithm RSS ^ and Radio Map * match, using the RSS ^ mean coordinates K nearest reference point as a test point (X ', 1'), which was expressed as:

Figure CN103648106AD00083

[0077] 式中,(X' , y')为测试点预测的坐标,(Xi, y)为第i个近邻点的坐标,K为KNN算法中近邻的数目; [0077] wherein, (X ', y') coordinates of the test point forecast, (Xi, y) coordinates of the i-th neighbor points, K is the number of neighbors KNN algorithm;

[0078] 四、用户定位终端定位数据库更新,即完成了基于类别匹配的半监督流形学习的WiFi在线阶段室内定位方法; [0078] Fourth, the user positioning terminal location database update, which completed the WiFi-based indoor positioning methods online stage semi-supervised manifold learning category matches;

[0079] 所述离线阶段在服务器上完成;在线阶段在定位终端上完成。 [0079] The stage is done on the server offline; online stages in the positioning of the terminal. [0080] 具体实施方式六:本实施方式与具体实施方式五不同的是:步骤三中对待定位点的具体位置坐标进行预测并进行在线数据的更新由权利要求1中所述的离线数据库和在线数据库实现: [0080] Specific embodiments of six: the present embodiment, five specific embodiment is different: Step three specific location coordinates to predict treatment positioning point and data is updated online by the claims in the off-line and on-line database of 1 database implementation:

[0081] 离线定位数据库的实现方式为:将定位用户本次定位测得的RSS作为未标记数据采用类别匹配方式加入到Radio Map中,并在移动终端上实现对本地在线匹配定位数据库的更新,实现动态的更新本地的数据,从而实现离线数据库定位方式; [0081] implementation of the database is offline Location: The target users of this positioning measured using RSS as unlabeled data matching category was added to the Radio Map in, and to implement local online matching updated location database on the mobile terminal, dynamic update local data, enabling offline database positioning mode;

[0082] 在线数据库的实现方式为:用户在线定位完成后,将用户本次在线测得的RSS值上传至在线定位数据库所在的服务器,并在服务器端将在线定位数据库进行更新将将在线定位数据传回上传RSS数据的定位终端。 Implementation [0082] online database is: users online positioning is completed, the user of this line measured RSS values uploaded to the online location server database resides, and the server side online location database updates will be online positioning data Upload RSS data back positioning terminal.

[0083] 其它步骤及参数与具体实施方式一至五之一相同。 [0083] one of one to five identical steps and other parameters specific embodiments.

[0084] 仿真实验: [0084] simulation:

[0085] 一、结合附图3对本仿真实验做出详细说明:图示为哈尔滨工业大学科学园2A栋12层的平面图示意,基于WiFi的室内定位系统就是基于该实验环境下建立。 [0085] First, with Figure 3 of the simulation to make a detailed description: Illustration of Harbin Institute of Technology Science Park, Building 12 layer 2A schematic plan view, WiFi-based indoor positioning system is based on the establishment of this experimental environment. 在实验环境中,总共布置27个AP,AP布置的位置为蓝色无线发射信号形状标记所在处。 In a lab environment, a total of arrangement 27 AP, AP is arranged at a location signal shape blue mark where the wireless transmitter. AP离房间地面高度为2米。 AP room on the ground from a height of two meters. 在离线阶段,在联想V450笔记本上安装NetStumbler软件,在所有参考点的四个不同的方位上连续采样记录AP的100个RSS值,以及AP的相关信息。 In the offline stage, NetStumbler software installed on Lenovo V450 notebook in four different directions all reference points continuous sampling record 100 RSS value AP, and AP information. 将所有的采样点的物理坐标及相应的物理坐标及RSS值存储为定位过程所调用的数据,建立Radio Map。 The data on the physical coordinates of all sampling points and the corresponding physical coordinates and RSS values are stored as locating the called procedure established Radio Map. 在实验环境共有900个参考点,其采样密度为0.5米X0.5米,如附图4所示。 A total of 900 experimental environment reference point, the sampling density is 0.5 m X0.5 m, as shown in Figure 4. Radio Map作为CM-SDE算法的输入参数及本征维数估计算法的输入数据。 Radio Map as an input parameter CM-SDE algorithms and intrinsic dimension estimation algorithm input data. [0086] 二、Radio Map的本征维数的获取通过下述步骤实现: [0086] Second, the intrinsic dimension of Radio Map is available through the following steps to achieve:

[0087] 本征维数是对于高维数据进行本征空间维数及空间重建所需最小的独立变量的个数。 [0087] intrinsic dimension of high-dimensional data for the intrinsic spatial dimension and spatial reconstruction of the required minimum number of independent variables. 在具体实际计算中,由于高维数据的本征并不明显,通常不是寻求得到确切的本征维数,而是寻求估计本征维数的可信取值。 In particular the actual calculation, due to the high dimensional data intrinsic is not obvious, it is not usually seek to get the exact intrinsic dimension, but seek credible estimate of the intrinsic value of dimension. 具体的说,给定一个来自高维空间的样本,本征维数估计算法的中心任务和重要内容就是通过这些样本数据来确定这个高维结构的本征维数。 Specifically, given a sample from a high-dimensional space, the intrinsic dimension estimation central task and an important part of the algorithm is through these sample data to determine the intrinsic dimension dimensional structure.

[0088] Radio Map的本征维数的估计是CM-SDE算法的重要输入参数,这关系到降维的结果是否能够代表Radio Map的高维空间的特征,因此准确有效的本征维数的估计至关重要。 [0088] Radio Map estimate of intrinsic dimension is an important input parameter CM-SDE algorithms, which relates to whether the results of dimensionality reduction can represent features Radio Map of high-dimensional space, so accurate and effective intrinsic dimension estimate is essential. 目前,常用本征维数估计算法分为两类:局部估计与全局估计。 At present, the common intrinsic dimension estimation algorithm is divided into two categories: local and global estimate estimate. 采用全局算法估计对RadioMap的本征维数进行估计,并作为CM-SDE算法的输入变量。 Global algorithms using an estimate of the intrinsic dimension RadioMap be estimated, and as input variables CM-SDE algorithms. 本实验中采用测地线最小生成树算法(Geodesic Minimum Spanning Tree, GMST)对Radio Map 的本征维数进行估计。 The experiment used a geodesic minimum spanning tree algorithm (Geodesic Minimum Spanning Tree, GMST) for intrinsic dimension Radio Map estimated.

[0089] 下面对GMST算法的理论进行分析。 [0089] The following GMST algorithm theory analysis.

[0090] 测地线最小生成树(GMST)估计是基于测地线最小生成树的长度函数依赖于本征维数d。 [0090] Geodesic minimum spanning tree (GMST) estimates are based on the length of the minimum spanning tree geodesic function depends on the intrinsic dimension d. GMST是指定义在数据集X上的近邻曲线的最小生成树。 GMST is a defined neighborhood graph in the dataset X on the minimum spanning tree. GMST的长度函数L(X)是在测地线最小生成树中所有边缘对应的欧氏距离之和。 GMST the length function L (X) is the geodesic minimum spanning tree all edges corresponding to the sum of Euclidean distance.

[0091] GMST估计在数据集X上构造一条近邻曲线G,其中,在X内每一个数据点Xi都和它的k个近邻"相连接。测地线最小生成树T定义为X上的最小曲线,它具有长度 [0091] GMST estimated curve G construct a neighborhood in the dataset X, wherein X within each data point are Xi and its nearest neighbors k "is connected geodesic minimum spanning tree T is defined as the minimum X on curve, which has a length

[0092] [0092]

Figure CN103648106AD00091

[0093] 其中,T是曲线G的所有子树集合,e是树T的一个边缘,ge是边缘e对应的欧氏距离,其计算公式为式(2)所示。 [0093] where, T is the graph G all subtrees set, e is an edge of the tree T, ge edge e corresponds to the Euclidean distance, which is calculated by the formula (2).

[0094] [0094]

Figure CN103648106AD00101

[0095] 在GMST估计中,一些子集Jc=X由各种大小m组成,并且子集A的GMST的长度 Length [0095] In GMST estimate, some subset of Jc = X m composed by a variety of sizes, and a subset of A of GMST

L(A)也需要计算。 L (A) can also be calculated. 理论上,f;是线性的,从而可以由y=ax+b这种形式的函数来估计, Theoretically, f; is linear, thus y = ax + b by the form of a function of this estimate,

维数的估计。 Estimate dimensions. 由GMST算法给出本征维数d的表达式为式(3)所示。 GMST algorithm given by the intrinsic dimension d of expression for the formula (3). 本征维数d是CM-SDE算法的另一个重要的输入参数。 Intrinsic dimension d is another CM-SDE algorithm important input parameters.

Figure CN103648106AD00102

[0097] 三、CM-SDE是一种半监督流形学习算法,在CM-SDE算法实现的过程中需要对所有的参考点进行类标记。 [0097] III, CM-SDE is a semi-supervised manifold learning algorithm, the process of CM-SDE algorithm requires a reference point for all class marks. 考虑到目前的WiFi室内定位环境中,参考点数目将近1000点,并没有人为地对所有的参考点进行类标记,而是采用一定的分类算法对参考点的类别进行标记。 Given the current environment WiFi indoor positioning, the reference point number nearly 1,000 points, and not artificially reference point for all the class mark, but the use of certain categories of classification algorithm reference point mark.

[0098] 聚类的目标是将数据集X= Ix1, X2,…,xj划分为c类且各类数据之间互不相关。 [0098] goal is to cluster the data set X = Ix1, X2, ..., xj into c classes and uncorrelated between various types of data. 基本的聚类算法按如下步骤实现: The basic clustering algorithm to achieve as follows:

[0099] (I)生成c个聚类中心,记为Vi, i=l,2,…,C。 [0099] (I) generate c cluster centers, referred to as Vi, i = l, 2, ..., C.

[0100] (2)将数据集X=U1, X2,…,xn}的每个元素归类,采用最近邻(Nearest Neighbor)算法判定元素的归属关系,其等价表达式为: [0100] (2) The data set X = U1, X2, ..., xn} of each element of the classification, nearest neighbor (Nearest Neighbor) algorithm attribution determination elements, the equivalent expression is:

[0101 ] [0101]

Figure CN103648106AD00103

[0102] 式(4)中,Xi为第i个数据点,Gi为第i类构成的邻接关系图。 [0102] formula (4), Xi is the i-th data point, Gi of the i-th class adjacency diagram configuration. D(Xi,Vj)表示计算Xi与'之间的欧式距离。 D (Xi, Vj) represents the Euclidean distance calculation Xi and 'between.

[0103] (3)聚类中心的更新,对于第i类的聚类中心更新为: [0103] (3) updating the cluster centers, the cluster centers class i updated to:

[0104] [0104]

Figure CN103648106AD00104

[0105] 式(5)中,II表示计算某类中元素的个数。 [0105] formula (5), II represents certain number of elements in the calculation.

[0106] (4)收敛性校验及迭代 [0106] (4) and iterative convergence check

[0107] 若满足以下四种情况下的收敛性条件之一,则迭代停止,否则重复执行(2)~(3),直到迭代收敛或者达到最大执行次数。 [0107] If one of the following four conditions for the convergence of the case is satisfied, the iteration stops, otherwise repeat (2) to (3) until the iteration convergence or the maximum number of executions. 四种收敛性判定条件为: Convergence of four kinds of conditions are determined:

[0108] 条件一:聚类中心不变; [0108] Condition One: cluster centers unchanged;

[0109] 条件二:每个聚类的元素不变; [0109] The second condition: Each element of the same cluster;

[0110] 条件三:聚类中心变化收敛于半径ε内; [0110] Three conditions: the cluster centers change in the radius of convergence ε;

[0111] 条件四:聚类元素变化收敛于半径ε内。 [0111] Conditions four: clustering changing elements converge within radius ε.

[0112] 总的来说,上述收敛性判定条件可以表述为下式: [0112] In general, the above-described convergence determination condition can be expressed by the following formula:

[0113] [0113]

Figure CN103648106AD00105

[0114] 其中,Vi'表示更新后的第i类的聚类中心。 [0114] where, Vi 'represents the first i kind of cluster centers after the update.

[0115] 从上述聚类算法的基本实现方式分析,用于判定元素的归属关系的算法构成了聚类的核心。 [0115] From the above basic implementation of clustering algorithm analysis algorithm for determining attribution elements constitute the core cluster. 不同类型的聚类算法提出不同的归类指标,一般将该函数称为损耗函数(LossFunction)。 Different types of clustering algorithm classify different indicators, generally the function is called loss function (LossFunction). 基本的聚类方法中采用欧式距离作为损耗函数。 The basic method used clustering Euclidean distance as the loss of function. 本专利采用KFCM算法对RadioMap进行类别分析得出聚类中心和Radio Map的类别标记。 This patent uses KFCM category analysis algorithms were derived RadioMap cluster centers and Radio Map of category tags. KFCM算法的理论分析如下: Theory KFCM algorithm is as follows:

[0116]引入核函数的模糊c均值聚类的目标是将原始的数据集所在空间变换至无穷维的希尔伯特空间(Hilbert Space),再对变换后的空间作相应的聚类分析。 [0116] The objective fuzzy c-means clustering kernel functions where the space is to transform the original data set to infinite dimensional Hilbert space (Hilbert Space), and then the space is transformed accordingly clustering analysis. 通过核函数的变换,将原始数据之间的类别特征进一步变换后更易于表述和区分。 By changing the kernel function, the easier to express and differentiate between the original data after class feature further transformation. 基于核函数的模糊c均值聚类算法的目标函数为: Fuzzy c-means clustering algorithm kernel function objective function is:

[0117] [0117]

Figure CN103648106AD00111

[0118] 式(7)中,Φ (xk)馮分别表示在希尔伯特空间下的数据集及相应的聚类中心。 [01] formula (7), Φ (xk) von respectively in the Hilbert space of data sets and the corresponding cluster centers. 通过推导可以得出KFCM算法的解表述为: Can be obtained by deriving solutions KFCM algorithm expressed as:

[0119] [0119]

Figure CN103648106AD00112

[0120] KFCM的解的关键在于计算希尔伯特空间的损耗函数或者相似度函数。 [0120] The key solution is to calculate KFCM loss function Hilbert space or similarity function. 在本文中考虑引入高斯核函数(Gaussian Kernel Function)的FCM (Fuzzy C-Means)算法的理论分析及其实现。 Theoretical analysis of algorithms and redistribute Gaussian kernel (Gaussian Kernel Function) of the FCM (Fuzzy C-Means) considered in this article. 高斯核函数如式(9)所示。 Gaussian kernel equation (9).

[0121] [0121]

Figure CN103648106AD00113

[0122] 希尔伯特空间中,由式表述其相应的损耗函数,该式进一步表述为式(10)。 [0122] Hilbert space, expressed by the formula corresponding loss of function, the formula is further expressed by Equation (10).

[0123] [0123]

Figure CN103648106AD00114

[0124] 式(10)中,〈.,.>表示计算相应式的核函数值。 [0124] (10), and <.,.> Represents the type of kernel functions to calculate the corresponding value. 而实际上,无穷空间的变换不存在,因此,对式(10)进一步简化为式(11)所示。 In fact, the transformation of infinite space does not exist, therefore, the formula (10) to further simplify (11) formula.

[0125] [0125]

Figure CN103648106AD00115

[0126] 式(11)的全展开式为式(12)所示。 [0126] Formula (11) The full expansion of the formula (12) below.

Figure CN103648106AD00121

[0128] 式(12)中,<Φ (xk), Φ (Xj) >由高斯核函数计算,即: [0128] formula (12), <Φ (xk), Φ (Xj)> calculated by the Gaussian kernel, namely:

[0129] [0129]

Figure CN103648106AD00122

[0130] 在算法实现中,不是随机生成聚类中心,而是从数据集X=Ix1, X2,…,χη}中随机选择c个元素作为聚类中心,构成集合Y= Iy1,…,y。 [0130] In the algorithm, the cluster centers are not randomly generated, but from the data set X = Ix1, X2, ..., χη} randomly selected as the cluster centers c elements constituting the set Y = Iy1, ..., y . }。 }. 因此,初始化的损耗函数值计算如式所示。 Therefore, the loss is calculated as a function of the value of the initialization formula.

[0131] [0131]

Figure CN103648106AD00123

[0132] 四、运用CM-SDE算法实现对Radio Map进行降维并获取特征权值矩阵过程通过下述步骤实现: [0132] Fourth, the use of CM-SDE algorithm to reduce the dimension of the Radio Map and get the weight matrix process characterized by the following steps to achieve:

[0133] CM-SDE算法是基于标记数据与未标记数据的类间散度及类内散度最大化的一种流形学习算法。 [0133] CM-SDE algorithm is based on the maximization of internal inter-class tag data and unlabeled data divergence and class scatter a manifold learning algorithms. 在对CM-SDE算法进行理论分析之前对CM-SDE算法给定的输入数据做如下 Before CM-SDE algorithm theory analysis of CM-SDE algorithm given input data as follows

说明:输入高维数据点匕}= e Rn,数据点Xi的类标记为Ji e {I, 2,…,P},其中P表示将高 Description: Enter the high-dimensional data point dagger} = e Rn, data point Xi class mark Ji e {I, 2, ..., P}, where P represents a high

维数据划分为P个子流形,即将输入的高维数据分成P类,记P类的聚类中心为V= Iv1, V2,…,VpI。 Dimensional data is divided into P sub-manifold, high-dimensional data input into the upcoming P class, remember P class cluster center is V = Iv1, V2, ..., VpI. 将输入的高维数据表示成矩阵的形式:X=[Xl,X2,…,xj e Rnxm0从矩阵表示的形式来看,矩阵中的列代表一个高维数据点。 High-dimensional representation of the input data in the form of a matrix: X = [Xl, X2, ..., xj e Rnxm0 from the point of view expressed in the form of a matrix, the matrix column represents a high-dimensional data points.

[0134] 对于包含未标记数据的RadioMapul,其中所有的未标记数据Xu= [xul, xu2,…,xj e Rnxk进行类别匹配,同时已标记数据记为X1=[Xll,X12,…,X1J GRn'对于Xu中的所有数据进行有序类别匹配。 [0134] For RadioMapul contain unlabeled data, wherein all of unlabeled data Xu = [xul, xu2, ..., xj e Rnxk conduct category match, but has been marked data recorded as X1 = [Xll, X12, ..., X1J GRn 'Xu all the data for the orderly category match. 有序的含义是当分配某一个未标记数据后,会影响相应类的聚类中心,因此会对下一未标记数据的类判别会有影响。 Ordered the meaning assigned one when unlabeled data, will affect the cluster centers corresponding class, so will the next class of data not marked discrimination would be affected. 本专利中主要的考虑信号的采集的时间顺序。 Chronological major concern in this patent acquisition signals. 假定Xu是按时间顺序排列。 Xu is assumed in chronological order. 采用式(15)计算类的归属Xul,并采用式(5)更新相应的聚类中心。 Using the formula (15) to calculate class home Xul, and using the formula (5) update the cluster center. 然后依次将所有的未标记数据进行类别匹配 Followed by all the unlabeled data category matches

[0135] [0135]

Figure CN103648106AD00124

[0136] CM-SDE算法的目标函数为: [0136] CM-SDE algorithm objective function is:

[0137] [0137]

Figure CN103648106AD00125

[0138] 式(16)中Sw、Sb、St分别表示类内散度、类间散度及总散度可以由式(17)计算:[0139] [0138] (16) in Sw, Sb, St denote the within-class divergence between class scatter and overall divergence can be represented by the formula (17) is calculated as follows: [0139]

Figure CN103648106AD00131

[0140] 式(17)中, [0140] formula (17),

Figure CN103648106AD00132

为第i类的均值,Ii为第i类的釆样点的数目; The mean first class i, Ii is the number of first class i Bian sample points;

Figure CN103648106AD00133

为全体釆样点的均值,N为釆样点的数目。 Bian is the mean of all sample points, N is the number of sample points Bian.

[0141] 对于式(16)所示的目标函数同样可以表示局部鉴别嵌入流形学习算法的目标函数形式,其表达式为: [0141] For the equation (16) shown in the objective function can also be embedded in the target in the form of a partial identification function manifold learning algorithms, and its expression is:

[0142] [0142]

Figure CN103648106AD00134

[0143] 式(18)中,Wij表示同类数据间的权重分配,Wi/表示不同类数据间的权重分配,分别表示为Wnxn和Wnxn'。 [0143] formula (18), Wij is the weight of similar data between redistribution, Wi / is the weight between the different types of data redistribution, denoted as Wnxn and Wnxn '. 权重计算过程由两步完成。 Weight calculation process is performed by two steps. 第一步:构造邻域图。 Step one: construct a neighborhood map. 根据高维数据点的类标记信息及其近邻关系构造无方向图G及G,。 Based on the class mark information and its neighbor relations point of high-dimensional data structure undirected graph G and G ,. 其中近邻关系是采用KNN算法给出的准则,即选择数据点最近的K个点作为其邻居,G表示当Xi与xj的类标记信息yi=yj时且X1、χj互为K近邻关系;G'示当Xi与χj的类标记信息Yi≠yj时且X1、Xj互为K近邻关系。 Wherein the neighbor relationship is the use of the guidelines given by KNN algorithm, choose the most recent data point K points as its neighbor, G represents the class when Xi and xj time stamp information yi = yj and X1, χj mutual neighbor relations K; G 'shows when Xi and Yi χj class stamp information when ≠ yj and X1, Xj K neighboring relationship to each other. 第二步:计算权值矩阵。 Step two: calculate the weight matrix. 根据第一步构造的邻接图采用类高斯函数进行权值矩阵的计算。 Calculate weighted matrix structure according to the first step in adjacency graph using Gaussian functions. 其表达式(19)、(20)为所示。 Its expression (19), (20) as shown in Fig. 公式中表示近邻点Xi与xj之间的权值,||X1-Xj ||2为近邻点.与xj之间的距离,采用矩阵方式计算距离,t为权值归一化参数,U、L分别表示未标记和已标记的采样点的数目。 The formula is the weight of neighbor points between Xi and xj, || X1-Xj || 2 points for the neighbors. And the distance between xj, using matrix calculation of the distance, t is the weight normalized parameters, U, L represent the number of untagged and tagged sample points. 根据分析可以知道,Wnxn和Wnxn'可以由三部分构成,分别是:已标记数据与已标记数据之间的权重、已标记数据与未标记数据之间的权重及未标 According to the analysis can know, Wnxn and Wnxn 'may be composed of three parts, namely: Marked data between the data that has been marked right weight, the right to data that has been marked with unlabeled data between heavy and not marked

记数据与未标记数据之间的权重,分别表示为 Right and unlabeled data recorded between weight data were expressed as

Figure CN103648106AD00135

[0144] [0144]

Figure CN103648106AD00136
Figure CN103648106AD00141

[0146]由上述计算公式及矩阵的性质可以得:WHz、Wlu=Wuxll、 [0146] from the nature of the above-mentioned formulas and matrices can be obtained: WHz, Wlu = Wuxll,

Wll =1 = 。 Wll = 1 =. 由此可以推导出Wnxn和Wnxn'表示为分块矩阵的形式,如式(21)所示。 It can be deduced Wnxn and Wnxn 'block is represented as a matrix form of equation (21).

Figure CN103648106AD00142

[0148] 根据矩阵的计算式: [0148] According to a matrix calculation formula:

Figure CN103648106AD00143

,计算式表示为矩阵A的矩阵的计算方法,计算式 The formula is expressed as calculated matrix A matrix formula

给出的方法与矩阵的迹的计算式一致,即:| |A| |2=tr(AAT)。 The method gives the trace of the matrix formula is consistent, namely: | | A | | 2 = tr (AAT). 由此式(18)可以表示为矩阵的迹的计算方式: Thus formula (18) can be expressed as the trace of a matrix calculation:

[0149] [0149]

Figure CN103648106AD00144

[0150] 式(22)可以简化为: [0150] (22) can be simplified as:

[0151] [0151]

Figure CN103648106AD00145

[0152]由矩阵迹的计算的标量性质及权值元素均为实数,可以将式(23)简化为: [0152] calculated by the matrix trace of scalar nature and weight elements are real numbers, can be of the formula (23) reduces to:

[0153] [0153]

Figure CN103648106AD00146

[0154] 根据简单的数学关系,可以将式(24)简化为: [0154] According to a simple mathematical relationship can be of formula (24) simplifies to:

[0155] J(V)=2tr{VT[X(D/ -Ψ NXN)XT]V} (25) [0155] J (V) = 2tr {VT [X (D / -Ψ NXN) XT] V} (25)

[0156] 式(25)中:X为输入数据,λ和V为特征值与特征向量,W和Wi分别为G及Gi对应的权值矩阵,D及D'均为对角阵,其对角元素可以由式(26)表示。 [0156] formula (25): X is the input data, λ, and V is the eigenvalues and eigenvectors, W and Wi G and Gi are right corresponding value matrix, D and D 'are diagonal, its diagonal elements may be represented by the formula (26).

[0157] [0157]

Figure CN103648106AD00147

[0158] 根据式(25)的推导方式,同理可以将式(18)中的约束条件写成如式(25)相似的形式,由此,可以将(18)表示为如下形式: [0158] According to equation (25) Derivation, empathy may be the formula (18) is written in a similar form of constraint equation (25), whereby you can (18) is expressed as follows:

[0159] [0159]

Figure CN103648106AD00151

[0160] 对式(27)应用拉格朗日(Lagrange)乘数法,可以得出式(28)所示: [0160] The formula (27) Lagrange (Lagrange) Multiplier Method, can be drawn (28) as follows:

[0161] [0161]

Figure CN103648106AD00152

[0162] 对式(28)进行广义特征值分解,得出其特征值分解的特征值及特征向量,表示为:A=LX1, λ 2,…,λ η]τ,其对应的特征向量为:V= [V1, V2,…,Vn] T。 [0162] The formula (28) generalized eigenvalue decomposition Eigenvalue Decomposition obtain its eigenvalues and eigenvectors, expressed as: A = LX1, λ 2, ..., λ η] τ, the corresponding eigenvector : V = [V1, V2, ..., Vn] T. 取前d个最大的特征值对应的特征向量构成变换矩阵V= [Vl,V2,…,vd]。 D take the first largest eigenvalue corresponding eigenvectors transform matrix V = [Vl, V2, ..., vd]. 由CM-SDE算法的输出数据变换方法可以得出,降维后数据为: By the output data conversion method CM-SDE algorithm can be drawn, low dimensional data:

[0163] [0163]

Figure CN103648106AD00153

[0164] 式(29)中,Zi表示输入高维数据点Xi变换后的低维输出数据。 [0164] formula (29), Zi represents the input low-dimensional output data high-dimensional data point Xi transformed. 从本专利中发明内容给出的离线阶段的实施步骤为:第一步先采用CM-SDE算法对所有参考点Radio Map进行降维处理,得到相应的参考点的降维后的Radio Map,即作为在线阶段的匹配定位数据库(RadioMap*)ο第二步对添加未标记数据的Radio Map,即RadioMapul进行降维处理,得到特征变换矩阵V'。 Implementation steps off stage content from this patent invention given as follows: The first step is the use of CM-SDE algorithm reference point for all Radio Map dimensionality reduction to give Radio dimensionality reduction corresponding reference point after Map, namely As the online phase matching location database (RadioMap *) ο second step unlabeled data Radio Map, namely RadioMapul dimensionality reduction, to give feature transform matrix V '. 由此可以建立离线阶段所需要数据库:RadioMap*和V'。 It can be set up off stage requires a database: RadioMap * and V '.

[0165] 五、由不同的用户在线阶段定位阶段获得的RSS是未标记类别属性的,其加入Radio Map,并构成Radio Mapul的过程称为类别。 [0165] Five different stages of positioning stage online users get the RSS is unlabeled category attribute, added Radio Map, and constitutes a process referred to as Radio Mapul category. 其实现方法如下所述: This is accomplished as follows:

[0166] 将不同用户在线阶段定位阶段测试得到的未标记RSS,采用类别匹配的方式加入至已有Radio Map中,得到相应的包含未标记信号覆盖图RadioMapul ;通过类别匹配方法增加Radio Map的数据量,进而提高Radio Map的密度,为CM-SDE算法提供新的降维数据,同时可以更新聚类中心,为CM-SDE算法提供新的类别数据。 [0166] The various stages of the positioning phase of the test users online obtained unlabeled RSS, using the category matching methods have been added to the Radio Map, to give the corresponding unlabeled coverage FIG RadioMapul comprising; increase Radio Map data by category matching method amount, thereby increasing the density of Radio Map, provides a new dimension reduction data for the CM-SDE algorithm, and can update the cluster center, offers a new category of data is CM-SDE algorithms. 类别匹配方法分为两步,其实现过程如下所述: Category matching method is divided into two steps, its implementation process as follows:

[0167] 第一步,寻找未标记RSS的类别属性。 [0167] First, look for the category attribute of unlabeled RSS. 记一组未标记RSS为RSSi,与步骤三中的聚类中心进行匹配,由式(4)完成RSSi的类别标记。 RSS remember a group is not marked as RSSi, and step three in the cluster center is matched by the formula (4) Complete RSSi category tags.

[0168] 第二步:对RSS进行门限检测。 [0168] The second step: the threshold for detection of RSS. 对于聚类中心vi表示为Vi=Gil, vi2,…,viN) ,N为室内定位系统中AP的个数。 For the cluster centers vi is expressed as Vi = Gil, vi2, ..., viN), N is the number of indoor positioning systems the AP. RSSi表示为RSSiKRSS^ RSSi2,…,RSSiN)。 RSSi expressed as RSSiKRSS ^ RSSi2, ..., RSSiN). 计算下式所定义的广义符号值: Defined by the following formula to calculate the generalized symbolic values:

[0169] [0169]

Figure CN103648106AD00154

[0170] 其中,sgn(.)定义为: (.) [0170] wherein, sgn is defined as:

[0171] [0171]

Figure CN103648106AD00155

[0172] 当Si大于设定门限值时,则将RSSi加入Radio Map中,并更新聚类中心,否则舍弃RSSi,不加入Radio Map中。 [0172] When the Si greater than the set threshold, then RSSi join Radio Map and updates the cluster center, otherwise reject RSSi, Radio Map not join in. 在本专利中,门限值VT=0.9N。 In this patent, the threshold VT = 0.9N. 聚类中心的更新由式(5)完成: Update the cluster centers by the formula (5) Completion:

[0173] 六、基于CM-SDE算法的WiFi室内定位方法的离线数据库实现方式: [0173] VI, algorithm-based CM-SDE offline database WiFi indoor positioning method is implemented:

[0174] 离线数据库方式由三部分构成。 [0174] offline database mode consists of three parts. 第一,所有参考点的Radio Map的建立,并采用CM-SDE算法得到RadioMap'第二,再随机采样U点未标记数据并添加入原有的Radio Map中,并用CM-SDE算法得到V',并将形成的定位数据库下载(存储)到定位的移动终端。 First, Radio Map of establishing all reference points, and the use of CM-SDE algorithm RadioMap 'second, then random sampling U unlabeled data point and added to the existing Radio Map and get V with CM-SDE algorithm' positioning database downloads and the resulting (storage) to locate the mobile terminal. 第三,在线定位实现及Radio Map更新。 Third, the online positioning to achieve and Radio Map Update. 第三部分的具体实现如下所述: The third part of the realization of concrete as follows:

[0175] 在线阶段,测试点处接收的RSS= [AP1, AP2,…,APn],η表示室内定位系统布置的AP的数目。 [0175] Live stage, received at the test points RSS = [AP1, AP2, ..., APn], η indicates the number of indoor positioning system layout of the AP. 将RSS与特征变换矩阵V'相乘,从而得出降维后的RSS' = [AP1, AP2,…,APJ,其中d表示本征维数。 The RSS and features of the transformation matrix V 'multiplied to arrive after dimensionality reduction of RSS' = [AP1, AP2, ..., APJ, where d is the intrinsic dimension. 再采用KNN算法实现RSS ^与RadioMap*的匹配。 Then using KNN algorithm RSS ^ match RadioMap * of. 采用与RSS^最近的K个参考点的坐标的平均值作为测试点(X',y'),其表达式为: RSS ^ mean coordinates using the nearest reference point K as a test point (X ', y'), which was expressed as:

Figure CN103648106AD00161

[0177] 将定位用户本次定位测得的RSS作为未标记数据采用类别匹配方式加入到RadioMap中,并在移动终端上实现对本地在线匹配定位数据库的更新,实现动态的更新本地的数据,从而实现离线数据库定位方式,附图1所示于类别匹配的半监督流形学习的WiFi室内定位方法在用户定位终端实现; [0177] The target users of this positioning measured using RSS as unlabeled data matching category was added to the RadioMap in, and to implement local Live match update locate database on mobile terminals, dynamically update data locally, thereby offline database localization, drawings 1 WiFi indoor positioning methods semi-supervised manifold learning in categories that match the user positioning terminal illustrated implementation;

[0178] 七、基于CM-SDE算法的WiFi室内定位方法的在线数据库实现方式: [0178] VII, CM-SDE algorithm based online database WiFi indoor positioning method is implemented:

[0179] 在线数据库方式由四部分构成。 [0179] online database mode consists of four parts. 第一,所有参考点的Radio Map的建立,并采用CM-SDE算法得到RadioMap'第二,再随机采样U点未标记数据并添加入原有的Radio Map中,并用CM-SDE算法得到V',并将形成的定位数据库下载(存储)到定位的移动终端。 First, Radio Map of establishing all reference points, and the use of CM-SDE algorithm RadioMap 'second, then random sampling U unlabeled data point and added to the existing Radio Map and get V with CM-SDE algorithm' positioning database downloads and the resulting (storage) to locate the mobile terminal. 第三,在线定位实现及Radio Map更新。 Third, the online positioning to achieve and Radio Map Update. 第三部分的具体实现如下所述: The third part of the realization of concrete as follows:

[0180] 在线阶段,测试点处接收的RSS= [AP1, AP2,…,APn],η表示室内定位系统布置的AP的数目。 [0180] Live stage, received at the test points RSS = [AP1, AP2, ..., APn], η indicates the number of indoor positioning system layout of the AP. 将RSS与特征变换矩阵V'相乘,从而得出降维后的RSS' = [AP1, AP2,…,APJ,其中d表示本征维数。 The RSS and features of the transformation matrix V 'multiplied to arrive after dimensionality reduction of RSS' = [AP1, AP2, ..., APJ, where d is the intrinsic dimension. 再采用KNN算法实现RSS ^与RadioMap*的匹配。 Then using KNN algorithm RSS ^ match RadioMap * of. 采用与RSS^最近的K个参考点的坐标的平均值作为测试点(X',y'),其表达式为: RSS ^ mean coordinates using the nearest reference point K as a test point (X ', y'), which was expressed as:

Figure CN103648106AD00162

[0182] 第四部分:用户在线定位完成后,将用户本次在线测得的RSS值上传至在线定位数据库所在的服务器,并在服务器端将在线定位数据库进行更新将将在线定位数据传回上传RSS数据的定位终端,即附图2所示的离线阶段在在线定位数据库所在服务器上完成,而在线阶段在定位终端完成。 [0182] Part IV: After users online positioning is completed, the user of this line measured RSS values uploaded to an online location server database resides, and the server will be updated online location database will be online positioning data back Upload positioning terminal RSS data, that the drawings shown in the offline phase completed locate the server where the database online, and the online phase is completed in positioning terminal.

Patent Atıfları
Alıntı Yapılan Patent Dosya kabul tarihi Yayın tarihi Başvuru sahibi Başlık
CN103079269A *25 Oca 20131 May 2013哈尔滨工业大学LDE (Linear Discriminant Analysis) algorithm-based WiFi (Wireless Fidelity) indoor locating method
US20110164522 *14 Mar 20117 Tem 2011Skyhook Wireless, Inc.Estimation of Position Using WLAN Access Point Radio Propagation Characteristics in a WLAN Positioning System
Patent Harici Atıflar
Referans
1 *邓志安: "基于学习算法的WLAN室内定位技术研究", 《哈尔滨工业大学博士学位论文》, 25 December 2012 (2012-12-25)
Referans veren:
Alıntı Yapan Patent Dosya kabul tarihi Yayın tarihi Başvuru sahibi Başlık
CN103906234A *3 Nis 20142 Tem 2014李晨Indoor positioning method based on WIFI signals
CN104185275A *10 Eyl 20143 Ara 2014北京航空航天大学Indoor positioning method based on WLAN
CN104469932A *21 Kas 201425 Mar 2015北京拓明科技有限公司Position fingerprint positioning method based on support vector machine
CN104507097A *19 Ara 20148 Nis 2015上海交通大学Semi-supervised training method based on WiFi (wireless fidelity) position fingerprints
CN104540221A *15 Oca 201522 Nis 2015哈尔滨工业大学WLAN indoor positioning method based on semi-supervised SDE algorithm
CN104581945A *6 Şub 201529 Nis 2015哈尔滨工业大学WLAN indoor positioning method for distance constraint based semi-supervised APC clustering algorithm
Sınıflandırma
Uluslararası SınıflandırmaH04W16/20, H04W64/00
Yasal Etkinlikler
TarihKodEtkinlikAçıklama
16 Nis 2014C10Entry into substantive examination
22 Mar 2017C14Grant of patent or utility model