阵列单通道高精度DOA估计与跟踪算法研究

发布时间：2018-08-15 16:05

【摘要】：阵列单通道相对于阵列多通道具有体积小、造价低以及系统复杂度低等优点。开展阵列单通道高精度空间谱估计技术研究对于测向系统的小型化应用具有重要意义。本文从阵列单通道的频域和时域模型入手,针对信号波达方向(Direction of Arrival,DOA)时不变和时变的情形,对信号DOA估计和跟踪问题进行了研究。本文的主要工作及成果如下:1.构建了阵列单通道的频域和时域空间谱估计模型。首先,在轮流采样结构下,将无限次轮巡得到的接收数据看作一个随机过程,一次轮巡得到的数据视为随机过程的一次实现(即样本函数),通过相位补偿建立了针对远场窄带信号的频域模型;然后利用信号在阵元间的延时不受其带宽影响这一特性,基于带通采样与重构定理,构建了延时补偿下的阵列单通道时域空间谱估计模型,使得宽、窄带信号可以统一到同一模型下处理。2.针对非均匀高斯白噪声背景,提出一种功率域最小二乘DOA估计器。首先推导了误差功率的最小二乘表达式,然后以此为目标函数,采用模拟退火遗传算法对其进行搜索优化,得到DOA估计值。模拟退火遗传算法既继承了遗传算法全局把握能力好的优势,又保留了模拟退火局部搜索能力强的特色,具有较高的搜索精度。此估计器给出了目标函数的闭式表达式,避免了现有算法的迭代求解及其造成的误差积累。并且由于表达式中不含与噪声功率相关的参数,因此所提估计器可以同时适用于均匀和非均匀高斯白噪声。3.针对阵列单通道下宽带信号DOA估计精度不高的问题,提出贝叶斯最大后验概率估计方法实现了阵列单通道下的宽带信号高精度DOA估计。首先推导了状态向量的后验概率密度函数,之后引入改进的完美抽样算法对其进行优化搜索,得到DOA估计值。由于状态向量的后验概率密度中包含其先验信息,并且与估计量无关的参数是通过积分消去而不是估计值替代,所以此方法可以获得较高的估计精度。4.对于信号方位缓慢变化的情形,将Cubature粒子滤波算法引入到阵列单通道DOA跟踪中。算法对基本粒子滤波中重要性密度函数的构造加以改进,将最新量测信息融入其中,使得从中抽取粒子的分布更加接近状态的后验概率密度,从而提高了DOA跟踪的精度。在递推过程中,为了解决误差协方差矩阵的对称性和正定性受到破坏而引起的数值不稳定问题,本文将平方根滤波与Cubature粒子滤波相结合,提出一种平方根Cubature粒子滤波(Square-root Cubature Particle,SCPF)算法,实现了阵列单通道下的高精度DOA跟踪。此算法通过对误差协方差阵取平方根保留了其对称性和正定性,并且减小了其条件数,提高了数值计算的稳定性。
[Abstract]:Single channel array has the advantages of small size, low cost and low complexity compared with multi-channel array. It is very important to study high-precision spatial spectrum estimation technology of single channel array for miniaturization of direction finding system. In the case of time-invariant and time-varying DOA, DOA estimation and tracking of signals are studied. The main work and achievements of this paper are as follows: 1. Estimation models of frequency-domain and time-domain spatial spectrum of array single channel are constructed. Firstly, the received data from infinite round trips are regarded as a random process, once, under the rotating sampling structure. The data obtained from the round trip is regarded as a random process (i.e. a sample function), and the frequency domain model of the far-field narrow-band signal is established by phase compensation. Then, based on the band-pass sampling and reconstruction theorem, the array single-channel time-space model with delay compensation is constructed by using the characteristic that the delay between the array elements is not affected by their bandwidth. In this paper, a power-domain least-squares DOA estimator is proposed for non-uniform Gaussian white noise background. Firstly, the least-squares expression of the error power is derived, and then the simulated annealing genetic algorithm is used to search and optimize the error power. The simulated annealing genetic algorithm (SAGA) inherits the advantage of good global grasp ability of GA and retains the characteristic of strong local search ability of SAGA, and has higher search precision. Moreover, the proposed estimator can be applied to both uniform and inhomogeneous white Gaussian noise due to the absence of noise power-dependent parameters in the expression. Degree DOA estimation. The posterior probability density function of the state vector is deduced firstly, and then an improved perfect sampling algorithm is introduced to optimize the search for DOA estimation. The Cubature particle filter algorithm is introduced into the array single channel DOA tracking in the case of slow signal azimuth change. The algorithm improves the structure of the importance density function in the basic particle filter, and incorporates the latest measurement information into it, so that the distribution of particles extracted from the array is better. In order to solve the problem of numerical instability caused by destroying the symmetry and positive definiteness of the error covariance matrix, a square-root Cubature particle filter (Square-Square-Square-Square-Square) is proposed by combining square-root filter with Cubature particle filter. The root Cubature Particle (SCPF) algorithm realizes high-precision DOA tracking in a single channel of the array. This algorithm preserves the symmetry and positive definiteness of the error covariance matrix by taking the square root, and reduces its condition number, thus improving the stability of the numerical calculation.
【学位授予单位】：解放军信息工程大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TN911.23

【参考文献】