基于动态射频指纹的调频定位方法  

Frequency modulation location method based on dynamic radio

作　　者：何艳军[1,2] 罗海勇[1] 戴永[2] 陈自力 

机构地区：[1]中国科学院计算技术研究所,北京100190 [2]湘潭大学信息工程学院,湖南湘潭411100 [3]北京三星通信研究院,北京100028

出　　处：《计算机应用》2014年第11期3173-3176,共4页journal of Computer Applications

基　　金：国家自然科学基金资助项目(61374214);国家科技重大专项(2014ZX03006003-002);国家863计划项目(2013AA12A201);太原市-中关村合作专项(130104)

摘　　要：针对调频(FM)广播信号动态变化对定位性能产生较大影响的问题,提出了动态射频指纹的FM定位方法。该方法采用多元线性回归和神经网络方法,根据离线阶段各个参考点和部分较准点的调频广播信号强度的关系,使用在线阶段较准点的信号强度实时估计在线阶段参考点的信号强度。通过这两种方法建立具有自适应能力的动态射频地图,并使用贝叶斯估计方法对目标进行定位。实验结果表明,相对静态射频指纹模型:采用多元线性回归的射频指纹动态映射模型定位误差平均减少9.1%,采用神经网络的射频指纹动态映射模型定位误差平均减少36.3%,有效抑制了射频信号动态变化时变性对定位性能的影响。In view of the fast that dynamic changes of Frequency Modulation（ FM） radio signal have great influence on the positioning performance, a localization method based on dynamic radio map was proposed. During offline phase, the relationship between radio signal strength of a few calibration points and each reference point was established by multiple linear regression and neural network, and then during online phase, the real-time radio signal strength values at all reference points were predicted based on the radio signal strength collected at calibration points in real-time. Dynamic map of radio frequency with adaptive ability was established using these two methods, and Bayesian estimation method was used to locate the target.The experimental results show that, compared with static radio map, the multiple linear regression model would reduce the deviation by 9. 1%, and the neural network model would reduce the deviation by 36. 3%, which restrains the negative impact of radio signal＇s time-dependent nature to the position performance.

关 键 词：调频广播信号 动态射频地图 多元线性回归 神经网络 

分 类 号：TP302.7[自动化与计算机技术—计算机系统结构] TP301.6[自动化与计算机技术—计算机科学与技术]