改进AdaBoost算法的WiFi室内定位  

WiFi indoor location based on improved AdaBoost algorithm

作　　者：贺超 吴飞[1] 张玉金 朱海 HE Chao;WU Fei;ZHANG Yujin;ZHU Hai(School of Electronic and Electrical Engineering,Shanghai University of Engineering Science,Shanghai 201620,China)

机构地区：[1]上海工程技术大学电子电气工程学院,上海201620

出　　处：《导航定位学报》2020年第5期32-36,共5页Journal of Navigation and Positioning

基　　金：国家自然科学基金项目(61272097);上海市自然科学基金项目(17ZR1411900);上海市科技学术委员会重点项目(18511101600);上海市信息安全综合管理技术研究重点实验室项目(AGK2015006)。

摘　　要：为了进一步研究现有自适应增强(AdaBoost)算法的无线保真(WiFi)室内定位方法中,指纹库数据异常值处理和子分类器的权重决策,提出1种改进AdaBoost算法的WiFi室内定位方法:通过1种判决式特征选择机制,优化特征属性的权重,减少指纹库数据异常值对子分类器的影响,有效提高子分类器的鲁棒性;在投票决策阶段,采用1种联合投票决策方法,充分保留对特征属性随机采样而导致的子树之间的多样性。实验结果表明,与已有相关算法相比,该算法能够有效降低训练阶段异常值对定位算法的影响,且分类准确率有显著提升。In order to further study on the handling of the abnormal data values for the fingerprint database and on the weight decision of the subclassifier in existing wireless fidelity(WiFi)indoor positioning methods of adaptive boosting(AdaBoost)indoor positioning methods of AdaBoost algorithm,the paper proposed a WiFi indoor location based on an improved AdaBoost algorithm:the weight of feature attributes was optimized and the influence of abnormal data values of the fingerprint database on the subclassifier was reduced to effectively improve the robustness of the subclassifier through an adjudicative feature selection mechanism;in the voting decision-making stage,a joint voting decision-making method was used to fully preserve the diversity among the subtrees caused by random sampling of the characteristic attributes.Experimental result showed that,compared with the existed algorithms,the proposed algorithm could effectively reduce the influence of abnormal values on the positioning algorithm during the training phase with the significant enhance of the classification accuracy.

关 键 词：室内定位 自适应增强算法 判决式特征选择 投票机制 基于位置的服务 

分 类 号：P228[天文地球—大地测量学与测量工程]