摘要
北斗导航信号在城市环境中容易受到复杂的干扰,这些干扰严重影响北斗接收机的定位精度。干扰特征中包含着对评估目的贡献较少的特征,这部分贡献较少的特征会影响分类结果的准确性。论文仿真了北斗卫星干扰信号并对信号进行特征提取组成预选特征集。同时提出了一种使用新的距离计算方式并可在所有样本核心圈进行迭代的改进型Re⁃liefF算法。可行性验证表明改进型ReliefF算法在选取特征为6维的情况下分类精度提升2%。最后使用改进型ReliefF算法提出预选特征集中贡献较大的部分,实现了由12维到8维的特征降维,得到8维最优特征子集。
Beidou navigation signals are susceptible to complex interference in urban environments,which seriously affect the positioning accuracy of Beidou receivers.The interference features contain features that contribute less to the evaluation purpose,and these features that contribute less will affect the accuracy of the classification results.In this paper,the Beidou satellite interfer⁃ence signal is simulated and the signal features are extracted to form a preselected feature set.At the same time,an improved Re⁃liefF algorithm that uses a new distance calculation method and can iterate in all sample core circles is proposed.The feasibility veri⁃fication shows that the improved ReliefF algorithm can improve the classification accuracy by 2%when the selected features are 6-dimensional.Finally,the improved ReliefF algorithm is used to propose the pre-selected feature set that contributes more,and the feature dimension reduction from 12-dimensional to 8-dimensional is realized,then the 8-dimensional optimal feature subset is gotten.
作者
刘蔚
王黎明
姚金杰
LIU Wei;WANG Liming;YAO Jinjie(Institute of Signal Capturing&Processing Technology,Key Laboratory of Shanxi Province,North University of China,Taiyuan 030051)
出处
《舰船电子工程》
2023年第10期82-86,131,共6页
Ship Electronic Engineering
关键词
北斗卫星导航信号
RELIEFF算法
特征降维
最优特征子集
Beidou satellite navigation signal
ReliefF algorithm
feature dimension reduction
optimal feature subset
