基于最小绝对收缩与选择算子模型稀疏恢复的多目标检测

发布时间：2019-07-07 21:27

【摘要】：针对地面多径环境下运动目标检测,使用最小绝对收缩与选择算子(LASSO)算法在参数估计时会出现伪目标的问题,提出一种基于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,使用宽带正交频分复用(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,在一定信噪比(SNR,-5 d B)下,基于设计矩阵降维构造方法的改进的LASSO算法比基追踪算法(BP)、DS(Dantzig Selector)、LASSO等传统算法的检测性能有明显提高;在一定虚警率(0.1)条件下,改进的LASSO算法比原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。
文内图片：

图片说明：多径场景示意图
[Abstract]:In order to solve the problem of pseudo-target in parameter estimation by using the minimum absolute shrinkage and selection operator (LASSO) algorithm in ground multi-path environment, a design matrix dimension reduction construction method based on LASSO model framework is proposed. Firstly, the multi-path propagation of the signal can bring the spatial diversity of target detection, and the signal has different Doppler frequency shifts on different paths. In addition, the use of broadband orthogonal frequency division multiplex (OFDM) signal can bring frequency diversity. The sparse characteristics of the target are caused by the introduction of spatial diversity and frequency diversity. The sparse vector is estimated by using the sparsity of multi-path and the prior knowledge of environment. The simulation results show that under a certain signal-to-noise ratio (SNR,-5 d B), the detection performance of the improved LASSO algorithm based on the design matrix dimension reduction method is significantly better than that of the traditional algorithm (BP), DS (Dantzig Selector), LASSO, and the detection probability of the improved LASSO algorithm is 30% higher than that of the original LASSO algorithm under the condition of a certain false alarm rate. The proposed algorithm can effectively remove false targets and improve the detection probability of radar targets.
【作者单位】：桂林电子科技大学广西无线宽带通信与信号处理重点实验室;桂林电子科技大学信息与通信学院;
【基金】：国家自然科学基金资助项目(61362006,61371107) 广西壮族自治区自然科学基金资助项目(2014GXNSFBA118288) 广西无线宽带通信与信号处理重点实验室基金资助项目(GXKL061501)~~
【分类号】：TN957.51

【参考文献】