【作者】 聂题；

【导师】 杨强；

【作者基本信息】 哈尔滨工业大学 ， 信息与通信工程， 2009， 硕士

【摘要】 空间谱估计具有超(高)的空间信号的分辨能力,因此也常常被称为“超(高)分辨谱估计”。本文首先讨论了空间谱估计的理论模型,然后分析了三种基于特征分解的线性预测算法(包括最大熵算法,最大方差算法以及自回归模型算法),之后对超分辨算法中最为经典的多重信号分类算法进行了深入研究。而这些算法都可以归结为子空间算法中的噪声子空间算法。子空间算法中另外一类就是旋转不变子空间算法。本文对基于最小二乘(LS)法和总体最小二乘(TLS)法的两类算法做出了在角度和成功概率等评价标准上的性能比较。上述算法均是针对方位的一维信号参数的估计。而同时对角度和多普勒频率的进行联合估计则更贴近实际。因此本文从一维参数估计的讨论出发,讨论了空域和时域处理等效性,并选用了噪声子空间类算法做了空时联合的扩展分析。在文中也详细的介绍了会影响多重信号分类(MUSIC)算法性能的因素。之后论文从计算机仿真出发,分析了阵元间距以及相干源信号对MUSIC的影响,并仿真验证了前后向平滑这种改进算法对相干信源有着良好的分辨能力。在实际海面,杂波会对超分辨算法的有效性造成很大的困扰。在文章中,特别采用高阶累积量的方法对MUSIC算法进行改进。这种改进不仅可以获得比二阶矩更好的性能,而且使得上述研究的超分辨算法无论在高斯白噪声环境还是在有色高斯噪声环境下均有很好的来波方向估计(DOA)性能。在文中,也基于这些改善,利用实测数据检验了MUSIC算法的性能。仿真是检验算法正确性的重要方法。论文中对每一种算法都做了仿真研究。在论文中,天线阵主要采用了线阵和面阵。研究结果表明,噪声子空间算法不仅在一维角度估计方面有良好性能,同时,通过对结果的定性分析,基于特征分解的噪声子空间算法完全能对二维参数(方位角和频率)进行联合估计。同时,本文选取了角度/频率绝对偏差,角度/频率方差以及成功概率对四种算法进行了定量分析。从比较结果来看,MUSIC从各个方面都优于其他算法。更多还原

【Abstract】 Spatial spectrum estimation is called super-resolution spectrum estimation for the reason that it has super (high) resolution for spatial signal. Based on the study of theoretical model, the dissertation first analyzes three linear prediction algorithms including maximum entropy algorithm based on character decomposition, the maximum variance algorithm and auto regression algorithm, and then studies multiple signal classification method, which is a classical super resolution algorithm. The entire above algorithms can be concluded into noise subspace algorithm.Another kind of subspace algorithm is estimation of signal parameters via rotational invariance techniques.This dissertation compares the algorithms of LS and TLS to evaluate the performance in the angle detection and success probability.All the above algorithms are one-dimensional signal parameters estimation aiming at Azimuth. However, 2-D estimation considering angle and doppler frequency is more applicable. This dissertation begins with the discussion of one-dimensional estimation, and discusses about the equivalent of time domain and spatial domain in dealing with signals. As to the above nosie space algorithms, the dissertation does further research on the space-time joint estimation.For the actual situation, the dissertation introduces the factors influencing the performance of MUSIC in detail and analyzes the effect of array element spacing and wideband signals on MUSIC. And it is verified that forward-back smoothing algorithm can distinguish coherent signals well.On real sea surface, clutter affects the resolution of the super resolution algorithm. At the end of the dissertation, higher-order statistics method is employed in improving MUSIC, which has better performance than second-order moment method, and it gains better DOA performance with super-resolution algorithm no matter in the environment of non-correlated Gaussian noise or correlated Gaussian noise.The dissertation simulates every algorithm, and selects linear array and area array as antenna array element. In this dissertation, real data are also employed to test the performance of MUSIC algorithm.From the simulation results, Noise subspace algorithm based on character decomposition can gain good performance in angle estimation and 2-D estimation for angle and frequency. As the test of the four algorithms, Absolute deviation and variance of angle/frequency and success probability are used to do quantitative analysis. From the result, MUSIC is better than other algorithms from all aspects.更多还原