穿墙雷达墙体与目标回波信号学习分离方法

卞粱; 晋良念; 刘庆华

引用本文:	卞粱，晋良念，刘庆华. 穿墙雷达墙体与目标回波信号学习分离方法[J]. 雷达科学与技术, 2022, 20(3): 289-296.[点击复制]
	BIAN Liang， JIN Liangnian， LIU Qinghua. A Learning Separation Approach for Wall Returns and Target Signals of Through-the-Wall Radar[J]. Radar Science and Technology, 2022, 20(3): 289-296.[点击复制]

【打印本页】【下载PDF全文】【查看/发表评论】【下载PDF阅读器】【关闭】

←前一篇|后一篇→

过刊浏览高级检索

本文已被：浏览 4241次下载 722次	码上扫一扫！
分享到：微信更多字体:加大+\|默认\|缩小-
穿墙雷达墙体与目标回波信号学习分离方法
卞粱，晋良念，刘庆华
1. 桂林电子科技大学信息与通信学院，广西桂林 541004;2. 广西无线宽带通信与信号处理重点实验室，广西桂林 541004

摘要:

在穿墙雷达成像中，事先有效分离墙体回波与目标信号，可以避免它们在后续的建筑物布局反演和内部目标成像中的相互影响。然而，现有的稀疏分离方法往往需要人工选择阈值参数，在一定程度上影响了分离效果，为此提出一种墙体回波与目标信号的学习分离方法。该方法将两信号的分离视为一种联合低秩-稀疏约束问题，使用迭代软阈值分离算法求解稀疏解，然后把稀疏解的迭代过程映射成多层神经网络中的每一层，并用数据集自适应训练所有层中的阈值参数。仿真和实测数据处理结果表明，该方法与人工选择阈值参数相比，有效提高了墙体与目标回波信号的分离效果。

关键词: 阈值参数联合低秩-稀疏约束迭代软阈值算法神经网络

DOI：DOI:10.3969/j.issn.1672-2337.2022.03.007

分类号:TN957.52

基金项目:国家自然科学基金(No.61861011, 61461012); 广西自然科学基金(No.2017GXNSFAA198050); 广西无线宽带通信与信号处理重点实验室2020年主任基金项目(No.GXKL06200106); 广西创新驱动发展专项(No.桂科AA21077008)

A Learning Separation Approach for Wall Returns and Target Signals of Through-the-Wall Radar

BIAN Liang， JIN Liangnian， LIU Qinghua

1. School of Information and Communication, Guilin University of Electronic Technology, Guilin 541004, China;2. Key Laboratorg of Guangxi Wireless Broadband Communication and Signal Processing, Guilin 541004, China

Abstract:

In through-the-wall radar imaging, the effective separation of the wall returns and target signals in advance can avoid their mutual influences on the subsequent building layout reconstruction and interior target imaging. However, the existing sparse separation methods often need to manually determine the threshold parameters in the algorithm, which affects their separation performance to a certain extent. In this paper, a learning separation method of the wall returns and target signals is proposed. The proposed method regards their se-paration as a problem of joint low-rank and sparsity constraint. The iterative soft threshold separation algorithm is used to solve the sparse solution. And then, the iterative process of this sparse solution is mapped to the multi-layer neural network. The threshold parameters of all layers are trained with data set adaptively. The simulation results and measured data show that the proposed method can improve the separation performance of wall returns and target signals compared with manual selection of threshold parameters.

Key words: threshold parameter joint low-rank and sparsity constraint iterative soft threshold algorithm neural network