基于线性收缩和随机矩阵理论的MIMO雷达目标盲检测方法
发布时间：2018-10-19 18:50
【摘要】：多输入多输出(MIMO)雷达作为一种新型雷达体制,已受到国内外学者的广泛关注。MIMO雷达由于能够有效地克服传统雷达体制存在的弊端,可以显著地提高目标的检测性能,具有巨大应用前景。目前,针对MIMO雷达目标检测已提出一系列方法,如Neyman-Pearson检测、广义似然比检测等,它们虽不同程度上提高了检测性能,但需要事先已知或预先估计噪声方差、目标散射矩阵等信息,属于非盲检测方法。而且,这些方法通常假定快拍数远远大于阵元数,此时,接收信号的样本协方差矩阵可以作为统计协方差矩阵的极大似然估计。随着MIMO雷达技术日益走向应用,大阵列系统已成为一个必然发展趋势。在大阵列系统中,阵元数可以与快拍数相比拟甚至大于快拍数,此时样本协方差矩阵的特征值分布区间发生改变,传统的目标检测方法不再适用。针对上述问题,本文以高维协方差矩阵的收缩估计技术和大维随机矩阵理论为工具,对MIMO雷达的目标盲检测方法进行了深入研究。本文的研究工作得到国家自然科学基金“基于大维随机矩阵理论的MIMO雷达稳健目标检测与估计”(项目编号:61371158)的资助。本文的创新性研究工作如下:针对阵元数与快拍数可以相比拟的大阵列MIMO雷达系统,将高维协方差矩阵估计的收缩算法与大维随机矩阵理论相结合,提出一种基于线性收缩-标准条件数(LS-SCN)的目标盲检测新方法。通过求解大维系统样本协方差矩阵的优化矩阵,并利用M-P律,推导出检测阈值与收缩系数之间的关系,分别给出了基于LS-SCN的单目标和多目标检测算法。该算法无需已知噪声方差、目标散射矩阵和目标方位等先验信息,对噪声变化不敏感,且适用于大阵列系统。针对快拍数相对于阵元数匮乏的情况,通过分析回波信号协方差矩阵的线性收缩系数的统计分布特性,提出一种基于收缩系数检测(SCD)的MIMO雷达多目标盲检测算法。进而,为了降低其计算复杂度,将收缩系数进行化简,选取特征值-矩之比(EMR)作为检测统计量,提出一种基于EMR的MIMO雷达多目标盲检测算法。仿真结果表明,两种算法显著地提高了MIMO雷达在快拍数匮乏环境下多目标盲检测的性能。传统目标检测方法通常仅考虑理想白噪声的情况,而实际中由于阵元间耦合等因素会产生相关噪声。针对此问题,本文建立了相关噪声模型,提出一种相关噪声背景下基于随机矩阵理论的MIMO雷达目标盲检测算法。该算法利用乘性自由卷积S-变换、加性自由卷积R-变换以及Stieltjes变换等数学工具,推导出其样本协方差矩阵特征值的渐近分布,结合标准条件数的检测思想,计算出其判决阈值,从而实现相关噪声背景下MIMO雷达的目标盲检测。
[Abstract]:As a new type of radar system, multi-input multi-output (MIMO) radar has attracted wide attention from scholars at home and abroad. MIMO radar can effectively overcome the disadvantages of traditional radar system and improve the performance of target detection. It has great application prospect. At present, a series of methods have been proposed for MIMO radar target detection, such as Neyman-Pearson detection, generalized likelihood ratio detection and so on. Although they improve the detection performance in varying degrees, they need to know or estimate the noise variance in advance. The target scattering matrix is a non-blind detection method. Moreover, these methods usually assume that the beat number is much larger than the number of matrix elements. In this case, the sample covariance matrix of the received signal can be used as the maximum likelihood estimation of the statistical covariance matrix. With the increasing application of MIMO radar technology, large array system has become an inevitable development trend. In large array systems, the number of array elements can be comparable to the number of beats or even larger than the number of beats. In this case, the distribution interval of eigenvalues of the sample covariance matrix is changed, and the traditional method of target detection is no longer applicable. In order to solve the above problems, the shrinkage estimation technique of high dimensional covariance matrix and the theory of large dimensional random matrix are used to study the blind target detection method of MIMO radar. The research work of this paper is supported by the National Natural Science Foundation of China "MIMO Radar robust Target Detection and estimation based on the large Dimension Random Matrix Theory" (item number: 61371158). The innovative research work of this paper is as follows: for large array MIMO radar systems where the number of array elements and rapid-beat numbers can be comparable, the contraction algorithm of high-dimensional covariance matrix estimation is combined with the theory of large-dimensional random matrix. A new blind target detection method based on linear contraction-standard condition number (LS-SCN) is proposed. By solving the optimization matrix of sample covariance matrix of large dimensional system and using M-P law, the relationship between detection threshold and shrinkage coefficient is derived, and the single object detection algorithm and multi-objective detection algorithm based on LS-SCN are presented respectively. The algorithm does not require prior information such as noise variance, target scattering matrix and target azimuth, so it is insensitive to noise changes and is suitable for large array systems. In view of the lack of rapid-beat number relative to the number of array elements, by analyzing the statistical distribution of linear shrinkage coefficient of echo signal covariance matrix, a blind MIMO radar multi-target detection algorithm based on shrinkage coefficient detection (SCD) is proposed. Furthermore, in order to reduce the computational complexity, the shrinkage coefficient is simplified and the ratio of eigenvalue to moment (EMR) is selected as the detection statistic. A blind detection algorithm for MIMO radar based on EMR is proposed. Simulation results show that the two algorithms can significantly improve the performance of MIMO radar blind detection in the absence of fast beat number. Traditional target detection methods usually only consider the case of ideal white noise, but in practice there will be correlation noise due to coupling between array elements. To solve this problem, a correlation noise model is established, and a blind detection algorithm for MIMO radar targets based on stochastic matrix theory is proposed. In this algorithm, the asymptotic distribution of eigenvalues of the sample covariance matrix is derived by means of multiplicative free convolution S- transform, additive free convolution R- transform and Stieltjes transform, and its decision threshold is calculated by combining the detection idea of standard condition number. In order to achieve the blind detection of MIMO radar under the background of correlated noise.
【学位授予单位】：吉林大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TN958
本文编号：2282047
[Abstract]:As a new type of radar system, multi-input multi-output (MIMO) radar has attracted wide attention from scholars at home and abroad. MIMO radar can effectively overcome the disadvantages of traditional radar system and improve the performance of target detection. It has great application prospect. At present, a series of methods have been proposed for MIMO radar target detection, such as Neyman-Pearson detection, generalized likelihood ratio detection and so on. Although they improve the detection performance in varying degrees, they need to know or estimate the noise variance in advance. The target scattering matrix is a non-blind detection method. Moreover, these methods usually assume that the beat number is much larger than the number of matrix elements. In this case, the sample covariance matrix of the received signal can be used as the maximum likelihood estimation of the statistical covariance matrix. With the increasing application of MIMO radar technology, large array system has become an inevitable development trend. In large array systems, the number of array elements can be comparable to the number of beats or even larger than the number of beats. In this case, the distribution interval of eigenvalues of the sample covariance matrix is changed, and the traditional method of target detection is no longer applicable. In order to solve the above problems, the shrinkage estimation technique of high dimensional covariance matrix and the theory of large dimensional random matrix are used to study the blind target detection method of MIMO radar. The research work of this paper is supported by the National Natural Science Foundation of China "MIMO Radar robust Target Detection and estimation based on the large Dimension Random Matrix Theory" (item number: 61371158). The innovative research work of this paper is as follows: for large array MIMO radar systems where the number of array elements and rapid-beat numbers can be comparable, the contraction algorithm of high-dimensional covariance matrix estimation is combined with the theory of large-dimensional random matrix. A new blind target detection method based on linear contraction-standard condition number (LS-SCN) is proposed. By solving the optimization matrix of sample covariance matrix of large dimensional system and using M-P law, the relationship between detection threshold and shrinkage coefficient is derived, and the single object detection algorithm and multi-objective detection algorithm based on LS-SCN are presented respectively. The algorithm does not require prior information such as noise variance, target scattering matrix and target azimuth, so it is insensitive to noise changes and is suitable for large array systems. In view of the lack of rapid-beat number relative to the number of array elements, by analyzing the statistical distribution of linear shrinkage coefficient of echo signal covariance matrix, a blind MIMO radar multi-target detection algorithm based on shrinkage coefficient detection (SCD) is proposed. Furthermore, in order to reduce the computational complexity, the shrinkage coefficient is simplified and the ratio of eigenvalue to moment (EMR) is selected as the detection statistic. A blind detection algorithm for MIMO radar based on EMR is proposed. Simulation results show that the two algorithms can significantly improve the performance of MIMO radar blind detection in the absence of fast beat number. Traditional target detection methods usually only consider the case of ideal white noise, but in practice there will be correlation noise due to coupling between array elements. To solve this problem, a correlation noise model is established, and a blind detection algorithm for MIMO radar targets based on stochastic matrix theory is proposed. In this algorithm, the asymptotic distribution of eigenvalues of the sample covariance matrix is derived by means of multiplicative free convolution S- transform, additive free convolution R- transform and Stieltjes transform, and its decision threshold is calculated by combining the detection idea of standard condition number. In order to achieve the blind detection of MIMO radar under the background of correlated noise.
【学位授予单位】：吉林大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TN958
【相似文献】
中国期刊全文数据库 前10条
1 李馨,陈晓苏,刘立刚;信息隐身术——信息隐藏的盲检测方法[J];计算机安全;2003年12期
2 刘蓉;霍甲;;信号盲检测应用情况简述[J];数字通信世界;2014年06期
3 张昀;于舒娟;王京;;基于自调节粒子群算法的盲检测[J];计算机技术与发展;2013年11期
4 王伟;方勇;;基于有限差分的置换图像盲检测方法[J];电子学报;2010年10期
5 刘潘梅;孙容海;吴建源;;一种新的区域复制图像篡改盲检测技术[J];计算机工程与应用;2012年09期
6 詹双环;张鸿宾;;基于小波分解和方差分析的图像信息隐藏盲检测[J];电子与信息学报;2007年06期
7 韩鹏;杨晓元;唐玉华;;基于一类支持向量机的隐秘图像盲检测算法[J];计算机工程与应用;2006年35期
8 平玲娣;刘祖根;史烈;孙康;;基于易变特征实现隐藏信息的盲检测[J];浙江大学学报(工学版);2007年03期
9 刘燕;刘朝阳;王安义;;一种快速传输格式盲检测的方法[J];数字通信;2011年03期
10 刘万贤;彭华;;一种突发直扩信号盲检测算法[J];信息工程大学学报;2013年06期
中国重要会议论文全文数据库 前2条
1 阮秀凯;张志涌;;Hopfield神经网盲检测统计信息缺失信号[A];2011年中国智能自动化学术会议论文集（第一分册）[C];2011年
2 罗向阳;王道顺;汪萍;刘粉林;;基于图像多域特征缩放与BP网络的信息隐藏盲检测[A];第七届全国信息隐藏暨多媒体信息安全学术大会论文集[C];2007年
中国博士学位论文全文数据库 前2条
1 胡玲娜;静止图像数字水印的盲检测算法研究[D];上海交通大学;2010年
2 吕志胜;基于ENF信号的数字音频篡改盲检测研究[D];华南理工大学;2014年
中国硕士学位论文全文数据库 前10条
1 吕晓辉;低截获概率信号的盲检测与参数估计[D];电子科技大学;2015年
2 年耀贞;宽带无线专网中PDCCH信道关键技术及性能研究[D];电子科技大学;2015年
3 宋啸良;量子蚁群优化盲检测系统设计[D];南京邮电大学;2015年
4 徐巧芬;面向图像高维隐写特征的盲监测算法[D];福州大学;2013年
5 季奎明;改进的Hopfield型神经网络盲检测算法研究[D];南京邮电大学;2016年
6 李垠;基于线性收缩和随机矩阵理论的MIMO雷达目标盲检测方法[D];吉林大学;2017年
7 夏yN;基于量子免疫优化的盲检测算法[D];南京邮电大学;2014年
8 张蓉;含公零点信道的信号盲检测[D];南京邮电大学;2012年
9 迟旭斌;直接序列扩频信号的盲检测和参数估计方法[D];西安电子科技大学;2014年
10 范乐园;半可逆FIR-MIMO系统多值信号盲检测[D];南京邮电大学;2014年
，本文编号：2282047
本文链接：https://www.wllwen.com/kejilunwen/xinxigongchenglunwen/2282047.html
