通信学报 ›› 2017, Vol. 38 ›› Issue (1): 158-167.doi: 10.11959/j.issn.1000-436x.2017018

• 学术通信 • 上一篇    下一篇

基于核函数特征提取的室内定位算法研究

李华亮1,钱志鸿1,田洪亮1,2   

  1. 1 吉林大学通信工程学院,吉林 长春 130012
    2 东北电力大学信息工程学院,吉林 吉林 132012
  • 修回日期:2016-10-14 出版日期:2017-01-01 发布日期:2017-01-23
  • 作者简介:李华亮(1990-),男,吉林长春人,吉林大学硕士生,主要研究方向为无线定位技术。|钱志鸿(1957-),男,吉林长春人,吉林大学教授、博士生导师,主要研究方向为基于物联网、D2D、Wi-Fi、RFID 等无线网络与通信技术。|田洪亮(1981-),男,吉林省吉林市人,吉林大学博士生,东北电力大学讲师,主要研究方向为无线个域网。
  • 基金资助:
    国家自然科学基金资助项目(61371092);国家自然科学基金资助项目(61401175);国家自然科学基金资助项目(61540022);吉林省科技发展计划基金资助项目(20140204019GX);长春市重大科技攻关计划基金资助项目(2014026/14KG021);吉林大学研究生创新基金资助项目(2019091)

Research on indoor localization algorithm based on kernel principal component analysis

Hua-liang LI1,Zhi-hong QIAN1,Hong-liang TIAN1,2   

  1. 1 College of Communication Engineering,Jilin University,Changchun 130012,China
    2 School of Information Engineering,Northeast Electric Power University,Jilin 132012,China
  • Revised:2016-10-14 Online:2017-01-01 Published:2017-01-23
  • Supported by:
    The National Natural Science Foundation of China(61371092);The National Natural Science Foundation of China(61401175);The National Natural Science Foundation of China(61540022);Scientific and Tech-nological Developing Scheme of Jilin Province(20140204019GX);Key Science and Technology Program of Changchun(2014026/14KG021);Project Supported by Graduate Innovation Fund of Jilin University(2019091)

摘要:

提出了一种基于核函数特征提取(KPCA,kernel principal component analysis)的室内定位算法。该算法在离线阶段使用核函数特征提取方法训练原始位置指纹(OLF,original location fingerprint),提取原始位置指纹的非线性特征,可以有效地利用各个接入节点(AP,access point)的接收信号强度信息;而在线阶段使用一种改进的加权k近邻 (IWKNN,improved weight k-nearest neighbor)算法,自主选择近邻数进行位置估计。实验结果表明,提出的算法在平均误差和定位准确率方面优于其他的室内定位算法,并且该算法需要更少的接收信号强度(RSS,received signal strength)采集次数和AP个数。

关键词: 无线局域网络, 室内定位, 接收信号强度, 核函数特征提取

Abstract:

An indoor localization algorithm based on kernel principal component analysis (KPCA) was proposed.It applied KPCA to train the original location fingerprint (OLF) and extract the nonlinear feature of the OLF data at the offline stage,such that the information of all AP was more efficiently utilized.At the online stage,an improved weight k-nearest neighbor algorithm for positioning which could automatically choose neighbors was proposed.The experiments were carried out in a realistic WLAN environment.The results show that the algorithm outperforms the existing methods in terms of the mean error and localization accuracy.Moreover,it requires less times of RSS acquisition and AP number.

Key words: WLAN, indoor localization, RSS, KPCA

中图分类号: 

No Suggested Reading articles found!