低复杂度阵列信号波达方向估计算法研究

发布时间：2018-01-12 22:37

本文关键词：低复杂度阵列信号波达方向估计算法研究　出处：《西安电子科技大学》2014年博士论文　论文类型：学位论文

【摘要】：阵列信号波达方向估计是阵列信号处理领域中具有重要应用价值的研究课题。近年来,低运算量、低复杂度的阵列信号波达方向估计算法受到了广泛关注。阵列信号波达方向估计将多个传感器设置在空间的不同位置组成传感器阵列,利用各个信号在空间位置上的差异,通过对多通道接收机输出数据的处理,获得空间各个信号源来波所对应的方向。在波达方向估计的发展历程中,基于传统理论和经典方法提出的有关算法既解决了一些实际问题,同时为该领域的发展提供了新的思路。由于波达方向估计的重要性使得该领域的研究经久不衰,并且对处理算法的性能提出了更高的要求。本文主要针对低运算量、低复杂度的阵列信号波达方向估计进行深入研究,主要研究成果有:1.针对传统基于子空间的波达方向估计算法中子空间估计运算量大的问题,提出了无需协方差矩阵特征值分解的快速波达方向估计算法。首先将子空间投影(SP)算法引入波达方向估计,将其与MUSIC算法相结合形成了子空间投影MUSIC算法;将子空间投影算法中大矩阵乘法运算用小矩阵乘积代替,提出了较低运算复杂度和存储量的简化子空间投影MUSIC算法;通过协方差矩阵分解简化子空间投影处理,得到低复杂度和存储量的高效子空间投影MUSIC算法。2.针对传统二维MUSIC算法中二维谱峰搜索运算量大的问题,提出只需一维谱峰搜索且无需角度配对的低复杂度二维波达方向估计算法。首先将降维MUSIC算法引入到二维波达方向估计领域,与求根法相结合,得到运算量更低的求根RD-MUSIC算法;将模值约束加入到RD-MUSIC算法的约束条件中,使方向向量的约束性更强,在保持低运算量的同时提高了角度估计性能;用直接求导法对RD-MUSIC算法进行求解,有效增强了方向向量中各元素应满足条件的约束性,同样在保持低运算量的同时进一步提高了角度估计性能。3.将阵列信号利用过完备基进行稀疏表示,可将波达方向估计问题转化成联合求解阵列信号在过完备基下的稀疏系数过程。提出两种基于单测量矢量的协方差矩阵稀疏重构波达方向估计方法,均无需已知信号数量,具有较低的运算复杂度和较高的角度估计性能。为了使算法具有相干信号估计能力,第一种算法采用了对协方差矩阵进行空间平滑处理的方式;第二种算法采用了构造协方差矩阵重构向量的方式。4.为了使稀疏重构波达方向估计方法只需少量样本就能达到较好的角度估计效果,采用了贝叶斯稀疏重构方法。介绍了多测量矢量稀疏贝叶斯学习(MSBL)算法和快速序列稀疏贝叶斯学习(FSBL)算法。分别将多种波达方向估计稀疏表示模型代入不同的贝叶斯稀疏重构算法进行仿真实验,并对多种不同算法的特点和性能进行了分析和总结。
[Abstract]:DOA estimation of array signals is an important research topic in the field of array signal processing. Low-complexity DOA estimation algorithms for array signals have attracted wide attention. The DOA estimation of array signals sets multiple sensors in different locations in space to form sensor arrays. By processing the output data of multi-channel receiver, the direction of each signal source is obtained. In the development of DOA estimation. The algorithm based on traditional theory and classical method solves some practical problems. At the same time, it provides a new way of thinking for the development of this field. Due to the importance of DOA estimation, the research in this field has not declined. And the performance of the processing algorithm is higher. This paper mainly focuses on the low computation, low complexity of array signal DOA estimation. The main research results are as follows: 1. Aiming at the problem that the traditional subspace-based DOA estimation algorithm has a large amount of computation. A fast DOA estimation algorithm without eigenvalue decomposition of covariance matrix is proposed. Firstly, the subspace projection algorithm is introduced into DOA estimation. The subspace projection MUSIC algorithm is formed by combining it with MUSIC algorithm. In this paper, a simplified subspace projection MUSIC algorithm with lower computational complexity and storage capacity is proposed, in which the large matrix multiplication is replaced by the small matrix product in the subspace projection algorithm. The subspace projection is simplified by covariance matrix decomposition. We obtain a high-efficient subspace projection MUSIC algorithm with low complexity and storage capacity. 2. Aiming at the problem of the large computation of two-dimensional spectral peak search in the traditional two-dimensional MUSIC algorithm. A low complexity 2D DOA estimation algorithm is proposed, which requires only one dimensional spectral peak search and no angle pairing. Firstly, the reduced dimension MUSIC algorithm is introduced into the field of 2D DOA estimation, which is combined with the root-finding method. The root finding RD-MUSIC algorithm with lower computation is obtained. The modular value constraint is added to the constraint condition of RD-MUSIC algorithm, which makes the direction vector more restrictive, and improves the performance of angle estimation while maintaining low computation load. The direct derivation method is used to solve the RD-MUSIC algorithm, which effectively enhances the constraint of each element in the direction vector. At the same time, the performance of angle estimation is improved. 3. The array signal is represented sparsely by over-complete basis. The DOA estimation problem can be transformed into solving the sparse coefficient process of array signals under overcomplete basis. Two methods of DOA estimation based on covariance matrix sparse reconstruction based on single measurement vector are proposed. In order to make the algorithm have the ability of coherent signal estimation, it has lower computational complexity and higher angle estimation performance without the need of known number of signals. In the first algorithm, the covariance matrix is processed by spatial smoothing. The second algorithm uses the method of constructing covariance matrix reconstruction vector. 4. In order to make sparse reconstruction DOA estimation method only a small number of samples can achieve a better angle estimation effect. The Bayesian sparse reconstruction method is used, and the multi-measure vector sparse Bayesian learning algorithm and the fast sequence sparse Bayesian learning algorithm (FSBL) are introduced. Algorithm. A variety of DOA sparse representation models are replaced by different Bayesian sparse reconstruction algorithms for simulation experiments. The characteristics and performance of different algorithms are analyzed and summarized.
【学位授予单位】：西安电子科技大学
【学位级别】：博士
【学位授予年份】：2014
【分类号】：TN911.7

【参考文献】