基于动态射频指纹的调频定位方法

Frequency modulation location method based on dynamic radio

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.