【作者】 徐东阳；

【导师】 叶中付；

【作者基本信息】 中国科学技术大学 ， 信号与信息处理， 2014， 博士

【摘要】 波束形成(Beamforming)技术是阵列信号处理领域中的一个重要研究方向,其利用空间多传感器所组成传感器阵列来对空间信号进行发射或接收,广泛应用于雷达,声呐,无线通信等领域。传统的波束形成技术假设信号是圆的、平稳的,仅仅利用了观测矢量的协方差矩阵,然而在通信领域大量的人工调制信号都是非圆的、非平稳的,不仅存在协方差矩阵而且存在伪协方差矩阵。基于圆信号假设下的波束形成技术虽然可以应用于非圆信号,但是其性能并不能达到最优。近些年现代信号处理领域中同时利用观测矢量及其共轭的宽线性(WL Widely Linear)技术的不断发展促使研究者们重新审视针对非圆信号的波束形成问题。宽线性波束形成技术通过分析信号的非圆性,同时利用了观测矢量的协方差矩阵和伪协方差矩阵,相应地一系列宽线性波束形成方法也陆续出现。本文在对宽线性波束形成技术进行概述的基础上,针对宽线性波束形成技术发展中存在的关键问题,开展基于非圆系数估计的宽线性波束形成算法研究。本研究工作的目的是在解决期望信号非圆系数估计问题的基础上,针对不同场景提出相应的宽线性波束形成算法。本论文研究工作的主要贡献与创新点归纳如下1.针对最优宽线性波束形成算法中非圆系数难以估计的问题提出非圆系数估计算法,并得到了一个稳健的对角加载权重矢量。我们研究使用最小功率无畸变(MPDR)准则,将其转化为一个参数估计问题,通过求导得到初始估计后,分析干扰和噪声对初始估计的影响,最后得到一个修正的估计值。在得到非圆系数估计后,使用对角加载技术减小估计误差对算法性能的影响。理论与仿真实验同时证明所提算法对非圆系数有比较好的估计性能,而且能有效地应用于实际场景。2.针对期望信号存在偏移频率的情况提出了频移宽线性波束形成算法。由于存在偏移频率的期望信号非圆系数随观测时间逐渐趋于零,共轭观测矢量与观测矢量中关于期望信号的相关信息是逐渐减少的。我们提出使用共轭循环观测矢量代替共轭观测矢量,其对应的共轭循环相关系数代替非圆系数进行宽线性波束形成。所提的算法在偏移频率为零时,等价于最优宽线性波束形成算法,在偏移频率不为零时,优于最优宽线性波束形成算法。此外针对期望信号是直线信号的情况,通过分析时间平均的非圆系数与偏移频率的关系,提出了两种估计偏移频率的算法。3.当期望信号的偏移频率与干扰的偏移频率不同时,提出M阶频移宽线性波束形成算法。频移宽线性波束形成算法仅利用了期望信号的共轭循环相关函数,当期望信号的偏移频率与干扰的偏移频率不同时,考虑使用多个共轭循环相关观测矢量,通过增加M个约束使其干扰在对应于他们自己的循环频率上的输出为零。所提算法是频移宽线性波束形成算法的扩展版本,不增加约束的零阶算法对应着频移宽线性波束形成算法。当期望信号的偏移频率与干扰的偏移频率不同时,所提算法优于频移宽线性波束形成算法。4.针对期望信号存在导向矢量失配的情况,引入非圆系数空间谱,提出了基于非圆系数空间谱估计的稳健宽线性波束形成算法。所提算法重构了干扰加噪声协方差矩阵和干扰加噪声伪协方差矩阵,剔除了扩展的协方差矩阵中的信号分量。在高信噪比和小快拍情况下,波束形成器输出信干噪比始终接近最优值。更多还原

【Abstract】 Beamforming technique is one of the most important issues in the array signal processing, which utilizes the array constituted by multiple sensors to transmit or receive the space signals, and has been widely used in the area of radar, sonar, communication and etc. Conventional beamforming technique generally assumes that the signals are circular and statistically stationary and only utilizes the covariance matrix of the observation vector. However, in the area of communication system, many artificial modulation signals are noncircular and nonstationary, which only have covariance matrix but also have conjugated covariace matrix. Although the conventional beamforming technique based on the circular assumption can be used for the reception of noncircular signals, the reception performance cannot attain the optimum. Recently, in the area of modern signal processing, the development of widely linear (WL) technique attains more and more researchers to reconsider the problem of the beamforming technique for the reception of noncircular signals. By analyzing the noncircularity of noncircular signals, the WL beamforming technique simultaneously utilizes the covariance matrix and conjugated covariance matrix of the observation vector and the corresponding WL beamforming methods were successively proposed.Based on the overview of WL beamforming technique, this paper aims at carrying out the study of the WL beamforming methods for the key issues in the development of WL beamforming technique. The target of our research is to solve the problem of noncircularity coefficient estimation and then propose corresponding WL beamforming methods under different scenarios. The main contributions and innovative points of this dissertation are listed as follows:1. For the problem of noncircularity coefficient estimation, we propose a method to estimate the noncircularity coefficient of the desired signal, and then a robust weight vector by using the diagonal loading technique. We consider the the principle of minimum Variance Distortionless response (MVDR) and turn out this problem into a parameter estimation problem. We find out the parameter to make the output power maximal as a rough estimate of noncircularity coefficient by taking the conjugate derivative of the objective function. Then, through analyzing the effect of interferences and noise on the rough estimate of noncircularity coefficient, we obtain the final modified estimate of the noncircularity coefficient. After obtaining the modified estimate of the noncircularity coefficient, the diagonal loading technique is used to suppress the effect of the estimation error on the optimal WL MVDR beamformer.2. For the situation that the desired signal has a non-null offset frequency, we propose a frequency-shift (FS) WL beamforming method for the reception of noncircular signals. Considering that the noncircularity rate of desired signal with a non-null offset frequency changes and approaches to0with the increment of observation time, the correlation information of the desired signal between the conjugated observation vector and observation decreases. We propose to use the conjugated cyclic observation vector and the corresponding conjugated cyclic coefficient instead of the conjugated observation vector and the noncircularity coefficient in the optimal WL beamformer. When the offset frequency of the desired signal is null. the proposed method is equivalent to the optimal WL MVDR beamformer. When the offset frequency of the desired signal is non-null.the proposed method outperforms the optimal WL MVDR beamformer. In addition, we propose two methods for the estimation of offset frequency based on relationship between the time-averaged noncircularity coefficient and offset frequency.3. Considering the situation that the offset frequencies of the desired signal and interferences are different, we propose an Mth-order FS WL beamfoming method. The FS WL beamfomer only utilizes the conjugated coefficient of the desired signal. When the offset frequencies of the desired signal and interferences are different, the proposed method uses multiple conjugated cyclic observation vectors and suppresses the output powers of interferences on their own cyclic frequency by adding M constraints. The proposed method is the extended version of the FS WL MVDR beamformer and the0th-order FS WL MVDR beamformer corresponds to the proposed FS WL MVDR beamformer. When the offset frequencies of the desired signal and interferences are different.the proposed method outperforms the FS WL MVDR beamformer.4. For the situation that there exists steering vector mismatch, we propose a robust WL beamforming method, in which the spatial spectrum of noncircularity coefficient is firstly introduced. The proposed method constructs the interfere-plus-noise covariance matrix (1NCM) and conjugated1NCM. Hence, the desired signal component is removed from the extended covariance matrix. Under the high signal-to-noise ratio (SNR) and small snapshots size, the output signal-to-interference-plus-noise ratio (SINR) of the proposed robust WL MVDR beamformer always approaches the optimal value.更多还原