阵列信号参数估计与跟踪方法研究

发布时间：2018-07-06 08:44

本文选题：阵列信号处理 + 信源数估计　；参考：《吉林大学》2016年博士论文

【摘要】：阵列信号参数估计与跟踪是阵列信号处理领域中一项重要的研究内容,其广泛应用于雷达探测、声纳定位、无线通信、地震勘探、射电天文以及生物医学工程等众多军事与民用领域。阵列信号参数估计与跟踪包括“参数估计”和“参数跟踪”两部分,其中阵列信号参数估计历经近几十年的发展,其基本理论框架和基本算法已经基本成熟,但随着阵列信号参数估计从理论方法到实际工程应用技术的转变,人们对算法的稳健性、估计精度以及计算复杂度等提出了更高的要求;而阵列信号参数跟踪更是面临着许多尚未得到很好解决的问题。本文深入研究了其中的一些关键问题,主要包括复杂噪声背景下的信源数估计、相干信号的波达方向(DOA)估计、相干/同向信号的DOA与多普勒频率联合估计、DOA跟踪以及多普勒频率跟踪等,针对以上问题提出了一系列有效的算法,并取得了一些有意义的成果。本文的研究工作主要包括以下6个部分:第一,针对复杂噪声背景下信源数估计问题,提出了一种基于伪协方差矩阵的Otsu类间方差法。由阵列输出信号延时相关构成的伪协方差矩阵对一定条件下的高斯白噪声和高斯色噪声具有免疫特性,并且利用阵元间的时间相关提高了阵列的有效孔径,进而提高了伪协方差矩阵奇异值分解后信号奇异值与噪声奇异值的大小差异程度。最后利用Otsu类间方差法对信号奇异值与噪声奇异值进行分类,进一步提高了信源数的检测成功概率。第二,针对相干信号DOA估计问题,提出了一种基于特征空间MUSIC算法的空间平滑估计方法。首先对相干信号进行空间平滑预处理,然后对其应用特征空间MUSIC算法进行DOA的有效估计,使其最大限度地利用信号子空间和噪声子空间的信息。所提方法并不影响非相干信号存在时DOA的估计,且还可以对信号源功率进行有效的估计。与传统空间平滑算法及修正MUSIC算法相比,所提方法具有更低的信噪比门限和更高的估计精度及分辨力。第三,针对DOA与多普勒频率联合估计问题,提出了两种算法:1)针对DOA与多普勒频率联合估计的二维MUSIC算法在低信噪比、小快拍数等非理想条件下估计性能变差的问题,提出了一种利用信号子空间投影方法修正的二维MUSIC算法。给出了一种运用信号功率倒数加权的信号子空间投影方法,并将其与二维MUSIC算法进行空间谱合成,充分利用信号子空间与噪声子空间两者的信息,提高了新算法在非理想条件下对多目标DOA及多普勒频率联合估计的分辨性能。2)针对相干/同向信号DOA与多普勒频率的联合估计问题,提出了一种基于数据共轭重构的修正二维MUSIC算法。首先,建立包含DOA与多普勒频率信息的广义阵列信号模型,并对阵列协方差矩阵利用数据共轭重排加以重构,使其可有效适用于相干/同向信号下DOA与多普勒频率的联合估计;其次,在二维MUSIC算法重构的基础上,利用1)中所提出的信号子空间投影方法对其加以修正,可以进一步提高算法的分辨性能。所提算法还可以对信号的功率进行有效估计,且估计参数均可自动配对。第四,针对最大似然估计方法计算量大的问题,提出了一种基于序列二次规划(SQP)的最大似然DOA及其与多普勒频率联合估计的方法。最大似然方法是一种在已知白噪声情况下的贝叶斯最优估计方法,在DOA、多普勒频率等参数估计问题中具有比特征子空间算法好得多的性能,并且它还可以直接处理相干信号。然而在其计算过程中需要非线性多维优化求解,针对传统网格搜索方法计算量过大的问题,提出了一种全局最优且局部具有超线性收敛特性的SQP方法,并把其应用于最大似然的优化求解中。最后通过仿真实验对其进行了有效性验证。第五,针对运动目标DOA跟踪问题,提出了一种时变遗忘因子的自适应样本协方差矩阵更新方法。时变遗忘因子根据DOA变化的快慢自适应调节自身的大小,从而合理地调整历史数据及当前采样数据在协方差矩阵更新过程中所占的权重;在协方差矩阵更新后,为了避免不断重复地进行特征值分解或奇异值分解,并且为了可以处理相干信号,对更新的协方差矩阵直接应用最大似然方法进行DOA估计。同时针对最大似然方法计算量较大的问题,分别利用人工蜂群仿生智能算法和SQP方法对其进行优化求解,有效减少了算法的运算量,加快了算法的优化速度,保证了跟踪的实时性。第六,针对雷达信号多普勒频率跟踪问题,提出了一种基于动态压缩感知的跟踪估计算法。首先建立雷达信号的稀疏时变信号模型,然后根据上一测量时刻稀疏向量中提取出的先验稀疏位置信息,构建当前时刻的冗余字典,并同时获得当前测量时刻稀疏向量中非零元素的分布概率,建立起多普勒频率的稀疏概率模型。最后,通过求解一个加权的l1范数最小化问题对当前稀疏信号进行重构,获得其非零元素位置,从而实现对多普勒频率的动态实时跟踪。仿真实验验证了所提算法的正确性与有效性。
[Abstract]:The estimation and tracking of array signal parameters is an important research content in the field of array signal processing. It is widely used in many military and civil fields, such as radar detection, sonar positioning, wireless communication, seismic exploration, radio astronomy and biomedical engineering. The estimation and tracking of array signal parameters include "parameter estimation" and "parameter". Following the development of the array signal parameter estimation in the past few decades, the basic theoretical framework and basic algorithms have been basically mature, but with the change of the estimation of the array signal parameters from the theoretical method to the actual engineering application technology, the robustness, the estimation accuracy and the computational complexity of the algorithm are higher than that of the two parts. A number of key problems are discussed in this paper, including the source number estimation under the complex noise background, the direction of arrival (DOA) estimation of coherent signals, the DOA of coherent / Homo signal and the Doppler frequency estimation, and the DOA tracking. And Doppler frequency tracking and so on, a series of effective algorithms are proposed and some meaningful results have been obtained. The research work of this paper mainly includes the following 6 parts: firstly, a Otsu inter class variance method based on pseudo covariance matrix is proposed for the source number estimation in complex noise background. The pseudo covariance matrix of the signal delay correlation is immune to the Gauss white noise and the Gauss color noise under certain conditions, and the effective aperture of the array is improved by the time correlation between the elements. The difference degree between the singular value of the pseudo covariance matrix and the singular value of the noise is improved. Then the Otsu inter class variance method is used to classify the singular value of signal and the singular value of noise, and the detection success probability of the number of sources is further improved. Second, in view of the DOA estimation problem of coherent signals, a spatial smoothing estimation method based on the feature space MUSIC algorithm is proposed. It uses the feature space MUSIC algorithm to estimate the DOA effectively, so that it can maximize the information of the signal subspace and the noise subspace. The proposed method does not affect the estimation of the DOA when the incoherent signal exists, and it can also effectively estimate the power of the signal source. Compared with the traditional spatial smoothing algorithm and the modified MUSIC algorithm, The proposed method has lower SNR threshold and higher estimation accuracy and resolution. Third, in view of the joint estimation of DOA and Doppler frequency, two algorithms are proposed: 1) the two dimensional MUSIC algorithm, which is combined with DOA and Doppler frequency estimation, is proposed to estimate the performance deterioration under low SNR, small snapshot number and other non ideal conditions. A two-dimensional MUSIC algorithm modified by the signal subspace projection method is presented. A signal subspace projection method using the weighted inverse of the signal power is given, and the spatial spectrum is synthesized with the two-dimensional MUSIC algorithm, and the information between the signal subspace and the noise subspace is fully utilized, and the new algorithm is improved in the non ideal condition. Target DOA and Doppler frequency joint estimation resolution.2) a modified two-dimensional MUSIC algorithm based on data conjugate reconstruction is proposed for the joint estimation of coherent / Homo signal DOA and Doppler frequency. First, a generalized array signal model containing DOA and Doppler frequency information is established, and the array covariance matrix is used. The data conjugate rearrangement is reconstructed so that it can be effectively applied to the joint estimation of the DOA and the Doppler frequency under the coherent / identical signal. Secondly, on the basis of the reconstruction of the two-dimensional MUSIC algorithm, the signal subspace projection method proposed in 1) can be used to improve the resolution of the algorithm. The power of the number is effectively estimated and the estimated parameters can be automatically paired. Fourth, a maximum likelihood DOA based on the sequence two times programming (SQP) and a joint estimation of the Doppler frequency are proposed for the maximum likelihood estimation. The maximum likelihood method is a kind of Bias with known white noise. The optimal estimation method has much better performance in the DOA, Doppler frequency and other parameter estimation problems, and it can also directly deal with the coherent signals. However, the nonlinear multidimensional optimization is needed in the calculation process, and a global optimum is proposed for the problem of too large computation in the traditional grid search method. The SQP method with superior and local superlinear convergence is applied to the optimal solution of maximum likelihood. Finally, the validity of the method is verified by simulation experiments. Fifth, an adaptive sample covariance matrix updating method for time-varying forgetting factor is proposed for the DOA tracking problem of moving target. According to the fast and slow change of the DOA, the weight of the historical data and the current sampling data in the covariance matrix update process is adjusted reasonably. After the covariance matrix is updated, in order to avoid repeated eigenvalue decomposition or singular value decomposition, and in order to process the coherent signal, the updated The covariance matrix uses the maximum likelihood method to estimate DOA directly. At the same time, aiming at the problem that the maximum likelihood method has large computational complexity, the artificial swarm intelligence algorithm and the SQP method are used to optimize the calculation, which can effectively reduce the computation of the algorithm, speed up the optimization speed of the algorithm, and ensure the real-time performance of the tracking. Sixth, the needle is guaranteed. For radar signal Doppler frequency tracking problem, a tracking estimation algorithm based on dynamic compression perception is proposed. Firstly, the sparse time-varying signal model of radar signal is set up, and then according to the prior sparse position information extracted from the sparse vector of the last measurement time, the redundant dictionary of the front time is constructed and the current measurement is obtained at the same time. The distribution probability of non zero element in the time sparse vector is established, and the sparse probability model of the Doppler frequency is established. Finally, by solving a weighted L1 norm minimization problem, the current sparse signal is reconstructed and its non zero element position is obtained, thus the dynamic real-time tracking of the Doppler frequency is realized. The simulation experiment verifies the proposed algorithm. Correctness and effectiveness.
【学位授予单位】：吉林大学
【学位级别】：博士
【学位授予年份】：2016
【分类号】：TN911.23

【相似文献】