计算机应用

• 人工智能与仿真 •    下一篇

基于LASSO模型稀疏恢复的多目标检测

洪刘根1,郑霖2,杨超1   

  1. 1. 桂林电子科技大学
    2. 桂林电子科技大学信息与通信学院
  • 收稿日期:2017-02-24 修回日期:2017-03-26 发布日期:2017-03-26 出版日期:2017-05-13
  • 通讯作者: 郑霖

Multi-target Detection via Sparse Recovery of the Lasso Model

  • Received:2017-02-24 Revised:2017-03-26 Online:2017-03-26 Published:2017-05-13

摘要: 摘 要: 针对地面多径环境下运动目标检测,使用LASSO算法在参数估计时会出现伪目标的问题,提出了一种基于LASSO模型框架的设计矩阵降维构造方法。首先,信号的多径传播能够带来目标检测的空间分集,信号在不同的多径上有不同的多普勒频移;此外,使用宽带OFDM信号能够带来频率分集。由于空间分集和频率分集的引入造成目标的稀疏特性。利用多径的稀疏性和对环境的先验知识,去估计稀疏向量。仿真结果表明,在一定虚警率条件下,基于设计矩阵降维构造方法的改进的LASSO算法比未改进的LASSO算法检测概率提高了30%。所提算法能够有效去除伪目标,提高雷达目标检测概率。

Abstract: Abstract: Focused on the issue that the lasso algorithm may introduce some false targets in the moving target detection in the presence of multipath reflections, the descending dimension method of the design matrix based on the lasso was proposed. Firstly, the multipath propagation increases the spatial diversity and provides different Doppler shifts over different paths. In addition, the application of broadband OFDM signal provides frequency diversity. The introduction of the spatial diversity and the frequency diversity to the system causes target space sparseness. The sparseness of multiple paths and the environment knowledge are applied to estimate paths along within receiving target responses. Simulation results show the target detection probability of the improved lasso algorithm based on the descending dimension method of the design matrix was 30% higher than the lasso in a certain false alarm rate. The proposed algorithm can effectively filter the false targets and improve the radar target detection probability.

中图分类号: