| 摘要: |
| 针对现有指纹定位技术存在指纹数据量较大、存储与处理困难、复杂空间定位适应性不足等问题，提出了一种基于大规模多输入多输出（Multiple-Input Multiple-Output，MIMO）系统的三维室内空间指纹定位方法。首先，提出一种处理速度更快、存储需求更小的角度-时延信道频率功率（Angle Delay Channel Frequency Power，ADCFP）指纹矩阵；其次，引入新的相似度准则即卡方距离以提高定位精度；然后提出一种改进的次方加权K近邻（Weighted K-Nearest Neighbor，WKNN）匹配算法，根据不同次方值对权重下降速度的影响差异，针对指纹相似度的大小分配以不同的权重；最后，对ADCFP指纹采用按行按列压缩的存储方法得到三种压缩指纹，进一步减少指纹数据量，并引入中心到达角（Central Angle of Arrival，CAOA）聚类算法缩短定位时长。仿真结果表明，ADCFP指纹矩阵2 m精度可靠性可达89.2%，采用卡方距离相较于曼哈顿距离的平均定位误差降低了5.63%，改进次方WKNN算法相较于传统WKNN算法平均定位误差降低了4.45%，引入CAOA聚类算法可使定位速度提升为未聚类情况下的1.72倍，平均定位误差较K均值聚类算法降低了44.05%，定位性能有较大提升。 |
| 关键词: 三维室内空间 指纹定位；大规模MIMO；加权K近邻(WKNN)；中心到达角（CAOA）聚类 |
| DOI：10.20079/j.issn.1001-893x.230831004 |
|
| 基金项目:国家重点研发计划（2022YFC38010000）；中央高校基本科研业务费专项资金（2242022k60001）；东南大学院系联合基金 （2242023K40015） |
|
| A Three-dimensional Space Fingerprint Localization Method Based on Massive MIMO System |
| HE Chenlin,WANG Xiaojun,WANG Lei |
| ((1.School of Information Science and Engineering,Southeast University,Nanjing 211189,China;2.Purple Mountain Laboratories,Nanjing 211111,China)) |
| Abstract: |
| In order to solve the problems of existing fingerprint localization technology,such as large amount of fingerprint data,difficulty in storage and processing,and insufficient adaptability to positioning in complex spaces,a three-dimensional indoor space fingerprint localization solution based on massive multiple-input multiple-output(MIMO) system is proposed.First,an Angle Delay Channel Frequency Power(ADCFP) fingerprint matrix with faster processing speed and smaller storage requirements is proposed.Secondly,a new similarity criterion,namely chi-square distance,is introduced to improve the positioning accuracy,and then an improved power Weighted K-Nearest Neighbor(WKNN) matching algorithm is proposed.The impact of the power value on the weight reduction speed is different,and different weights are allocated according to the fingerprint similarity.Finally,three types of compressed fingerprints are obtained by using row-by-column compression of ADCFP,further reducing the amount of fingerprint data.And the Central Angle of Arrival(CAOA) Clustering Algorithm is introduced to shorten the positioning time.The simulation results show that the ADCFP fingerprint matrix can offer a 89.2% reliability for 2 m accuracy.The average positioning error using chi-square distance is reduced by 5.63% compared with that using the Manhattan distance.The improved power WKNN algorithm reduces the average localization error by 4.45% compared with the traditional WKNN algorithm.The introduction of CAOA Clustering Algorithm can increase the localization speed to 1.72 times that of the non-clustering case.The average localization error is reduced by 44.05% compared with the K-means Clustering Algorithm,and the positioning performance is greatly improved. |
| Key words: 3D indoor space fingerprint localization massive MIMO weighted K-nearest neighbor(WKNN) central angle of arrival(CAOA) clustering |
