\r刘开华，吕 粮，马永涛\r
\r
AuthorsHTML:\r刘开华，吕 粮，马永涛\r
\r
AuthorsListE:\rLiu Kaihua，Lü Liang，Ma Yongtao\r
\r
AuthorsHTMLE:\rLiu Kaihua，Lü Liang，Ma Yongtao\r
\r
Unit:\r天津大学微电子学院，天津 300072\r
\r
Unit_EngLish:\rSchool of Microelectronics，Tianjin University，Tianjin 300072，China\r
\r
Abstract_Chinese:\r\r基于指纹的超高频射频识别\r(\rUHF RFID\r)\r定位现存的方法主要依赖于一些单指纹和全局参考点上，针对指纹库存在数据量冗余和易受到环境变化影响的问题，提出了一种方位优选多信息指纹\r(\rOPMIF\r)\r技术．\rOPMIF\r是一种指纹融合技术，它结合对目标到达角\r(\rDOA\r)\r的模糊估计，为不同位置目标构建特定的指纹数据库．首先，通过阵列天线相位控制筛选方位优选参考标签\r(\rOPRT\r)\r，无需精确计算目标的\rDOA\r因而不会增加计算量和定位时延．然后，利用阵列天线接收\rOPRT\r标签的多频多径响应\r(\rMFMP\r)\r获取多种指纹．多种指纹融合包含更多环境相关的信息，可以改善单一指纹定位易受环境变化影响的缺点．多信息指纹\r(\rMIF\r)\r包括信号协方差矩阵\r(\rSCM\r)\r、接收信号强度\r(\rRSS\r)\r和信号子空间\r(\rSSP\r)\r．然后将定位问题转化为模式识别问题，基于集成学习算法随机森林\r(\rRF\r)\r设计多个分类器来训练不同类型的指纹．最后，针对同一样本数据多指纹训练的不同学习器的估计结果可能不同，同时对于不同采样数据的同一种指纹学习器的估计结果也有可能不同，提出了一种后验权重估计定位\r(\rPWEL\r)\r算法来融合不同分类器和样本的预测．仿真实验通过定位误差累积分布函数\r(\rCDF\r)\r曲线来对比单指纹与多指纹融合，全局指纹与方位优选指纹的性能来验证\rOPMIF\r的有效性，并在真实场景中做实验验证提出的定位系统性能优于传统的单指纹方法\r.\r\r
\r
Abstract_English:\r\rThe existing fingerprint-based UHF RFID localization approaches rely mainly on some single and global fingerprint reference points\r．\rTo address the problems of the fingerprint library\r，\ri.e.\r，\rdata redundancy and vulnerability to environmental changes\r，\ran orientation priority multi-information fingerprint\r(\rOPMIF\r)\rwas proposed\r．\rThe OPMIF is a fingerprint fusion technique that combines the fuzzy estimates of the direction of arrival\r(\rDOA\r)\rto build specific fingerprint databases for different location targets\r．\rFirst\r，\rorientation priority reference tags were selected by controlling the phase of the array antenna\r，\rwhich eliminates the need to accurately calculate the DOA of the target and does not increase the amount of computation and localization delay\r．\rThen\r，\rmultiple fingerprints were obtained by the array antenna receiving the multifrequency multipath response\r．\rMultiple fingerprint fusions contain more context-related information that can improve the vulnerability of single fingerprint localization to environmental changes\r．\rThe multi-information fingerprint includes the signal covariance matrix\r，\rreceived signal strength\r，\rand signal subspace\r．\rSubsequently\r，\rthe localization problem is transformed into a pattern recognition problem\r．\rMoreover\r，\rmultiple types of classifiers are designed on the basis of an integrated learning algorithm\r(\ri.e.\r，\rrandom forest\r)\rto train different types of fingerprints\r．\rFinally\r，\rthe estimation results of different learners for the same sample data may be different\r．\rAt the same time\r，\rthe estimation results of the same learner for different sampling data may be different\r．\rThus\r，\ra posterior weight estimation localization algorithm was proposed to combine the predictions of different classifiers and samples\r．\rThe simulation experiments compare the performance of single fingerprint versus multiple fingerprint fusion and global fingerprint versus preferred orientation fingerprint using the cumulative distribution function curves of the localization errors to verify the validity of the OPMIF\r．\rMoreover\r，\rthe experiments in real scenarios verify that the proposed localization approach is superior to the traditional single fingerprint methods\r．\r\r
\r
Keyword_Chinese:射频识别；方位优选多信息指纹；方位优选参考标签；多频多径响应；后验权重估计定位算法\r

Keywords_English:radio frequency identification(RFID)；orientation priority multi-information fingerprint(OPMIF)；orientation priority reference tag(OPRT)；multifrequency multipath response(MFMP)；posterior weight estimation localization(PWEL) algorithm\r


PDF全文下载地址:http://xbzrb.tju.edu.cn/#/digest?ArticleID=6415