【作者】 甘露；

【导师】 魏平；

【作者基本信息】 电子科技大学 ， 信号与信息处理， 2009， 博士

【摘要】 分析与识别日益复杂的通信信号以及对其辐射源测向在军事以及民用领域中都有着非常重要的作用。确定信号的波达方向是电子侦察系统的一个重要方面。频率估计和频谱分析是在电子侦察系统中进一步分析复杂通信信号的基础。混沌直扩信号具有抗侦察、抗截获能力，将会在未来军事通信中发挥越来越重要的作用，对其的盲分析技术是近年来电子侦察中的研究热点之一。本文的主要工作是对通信信号电子侦察系统进行研究：针对二维波达方向估计中的高精度参数估计与配对问题，利用特定的阵列结构和信号时域特征开展研究；针对频域分析技术，研究了几种频率估计算法和频谱分析方法；最后分析了实值混沌直扩系统的盲解扩技术。本文的主要创新之处概括如下：针对单L阵二维波达方向估计中的配对算法开展了深入研究。提出了利用联合奇异值分解，实现了方位角与俯仰角的自动配对，同时提高了测向精度；两种改进互相关配对算法解决了原算法中对测向范围的限制；利用时间信息构成伪协方差矩阵，能实现到达角与频率的分别估计，最后通过伪协方差矩阵的互相关实现到达角与频率的配对。针对双L阵提出一种二维DOA估计算法，该算法利用了这种特殊的阵列结构形式所形成的空间角度约束，使用简单的算法即可实现方位角与俯仰角的配对；利用空间角度约束还能采用加权最小二乘算法来进一步提高角度估计精度，并给出了三种实现方案。针对单L阵提出了两种利用信号非圆特性的二维DOA估计算法。联合SVD算法同时通过利用非圆信号特征以及L阵的结构，能适应更多信号源个数，实现更高精度测向以及自动配对。快速算法由于利用了信号的非圆特性比常见的快速算法，如传播因子算法(PM)等具有更高的测向精度。针对频率估计技术提出了三种快速频率估计技术。多通道迭代的测频算法，具有测频精度高，测频信噪比门限低，具有易于硬件实现的特点。基于部分自相关技术的测频算法，比常用的基于自相关技术的频率估计方法计算量更小。基于前置滤波的单频信号估计算法的估计精度与待测信号频率无关，适合于电子侦察接收机使用。针对频谱分析，提出了一种非均匀傅里叶变换的快速计算算法，能实现对任意频谱上的频率分量的准确、快速计算，适合于电子侦察接收机使用。针对短码实值混沌直扩系统，提出了一种基于主元分析的算法，实现了在低信噪比下对直扩序列的盲估计。针对长码实值混沌直扩系统，提出了两种盲解扩算法。非线性建模预测的算法，实现了对长码实值直扩序列的盲解扩，以及扩频增益与信息码的同步等参数的估计；在已知上述参数时，可以利用混沌广义同步与无先导kalman滤波的方法来实现对长码实值直扩序列的盲解扩，该算法能在较低信噪比的情况下获得良好的性能。更多还原

【Abstract】 It is very important to analyze and identify the complex communication signals, and also to locate the impinging sources in the military and civilian areas. Direction of arrival (DOA) estimation is an important topic in electronic reconnaissance system. Frequency estimation and spectrum analysis are the foundation of the analysis of complex communication signals in electronic reconnaissance system. Chaotic direct sequence spread spectrum (DSSS) communication systems have merits of anti-reconnaissance and low probability of intercept, and they will play an increasingly important role in future military communications. Blind estimation of chaotic DSSS signal has received considerable attention in recent years.The main work of this dissertation is studying the electronic reconnaissance system: (1) The structure of the array sensors together with the signal properties in time domain are used to study the problems of high-resolution two-dimensional (2-D) DOA estimation and pair-matching; (2) For spectrum analysis, several frequency estimation algorithms and spectrum analysis methods are developed; (3) Blind estimation techniques of chaotic DSSS signal are studied. The main creative works are concluded as follows.We make an in-depth study on pair-matching in joint estimation of 2-D DOA for L-shape arrays. A new method of 2-D DOA estimation based on joint SVD is proposed to achieve automatically pair-matching with better estimation precision. Furthermore, two modified pair-matching algorithms based on the cross-correlation matrix are developed, which extend the range of the elevation and azimuth angles. An algorithm of joint estimation frequency and DOA based on the pseudocovariance matrix in time domain is investigated. The DOA and frequency are estimated respectively, and the pair-matching can be achieved with the cross-correlation of pseudocovariance matrices.A new algorithm of joint estimation of 2-D DOA for double L-shape arrays is investigated. This method makes use of a restriction of space direction formed by the special structure of array sensors. The new method can not only be used to pair the elevation angles and azimuth angles, but also improve the estimation accuracy by weighted least squares algorithm. Three implementation schemes are given at last.Two methods of 2-D DOA estimation for L-shape arrays with noncircular sources are proposed. The method of joint SVD can be adaptive to more signal sources, achieve a better estimation precision and automatically pair 2-D angle estimates by the merits of noncircular characteristics and the structure of L-shape arrays. By using the noncircular characteristics, the fast algorithm provides a better performance than the traditional fast algorithm such as propagator method (PM).Three fast frequency estimation techniques are discussed respectively. The iterative method of frequency estimation based on multi-channels, with high estimation precision, low SNR threshold, is easy to implement with hardware. The method based on partial auto-correlation technique is computationally efficient. And the estimation precision of the algorithm based on pre-filtering does not depend on the signal frequency to be estimated. For spectrum analysis, a fast algorithm for non-uniform Fourier Transform is developed to fast compute the signal spectrum at arbitrary frequency components, and it can be used in the receiver of electronic reconnaissance.A new method of blind estimation for short-code chaotic spread spectrum sequence is proposed, which is based on principal component analysis (PCA) and has good performance at low SNR condition. Two new methods are developed for blind estimation for long-code chaotic spread spectrum sequence. One is based on nonlinear modeling and prediction, and it can despread the long-code chaotic DSSS signal blindly and obtain the parameters, such as spread gain and information code synchronization. The other is based on the generalized synchronization (GS) and Unscented Kalman filter (UKF), and it can despread the long-code chaotic DSSS signal blindly and has a good performance at low SNR condition.更多还原