【作者】 张辉；

【导师】 葛临东；

【作者基本信息】 解放军信息工程大学 ， 信号与信息处理， 2007， 博士

【摘要】 本文致力于二维空间谱估计和自适应波束形成技术研究,本文所作的工作是实验室承担的某重点预研项目的一部分。作为阵列信号处理的主要研究内容,波达方向(DOA)估计和波束形成(BF)得到了人们广泛的关注,经过多年的研究获得了大量的理论成果。与一维相比,二维DOA估计和波束形成更具普遍意义和实用价值,它能够对空间信号实现更准确的定位,在通信中也能更加充分的利用信号的空间特征提高系统容量和通信质量。然而,针对二维参数的阵列信号处理有其特殊性,人们对二维空间谱估计和自适应波束形成技术的研究还不充分,还存在不少问题没有得到很好的解决,比如角度兼并问题、阵元利用率问题、瞬时信号或快速运动信号的DOA估计问题、多用户多径信号DOA估计问题、稳健的自适应波束形成问题等,这也是人们研究的部分热点和难点问题。本文结合某重点预研项目,针对以上这些方面进行了深入的研究,提出了一些新的或改进的方法,进一步完善和丰富了阵列信号处理理论。分析了波达方向矩阵类算法存在的角度兼并问题,并提出了改进方法,使这类算法更加完善。并指出角度兼并问题广泛地存在于以旋转不变结构为基础的二维或多维参数估计算法中,凡是通过估计方向向量并以此求解参数的算法都无法避免角度兼并,只能在发生兼并时采取一些辅助手段加以解决。分别针对双平行均匀线阵和均匀圆阵提出了基于时间平滑技术的多用户多径二维DOA估计方法。该方法适用于移动通信环境,要求信道满足块衰落模型,能够对高斯噪声中的非高斯信号进行二维DOA估计,并且能对来波按相干性实现分组。这种算法能够估计的来波数目远远超过了阵元数目,是现有其它算法无法实现的。但该方法运算量大,对处理器速度要求较高。首次提出了一种单快拍二维DOA估计算法——单快拍DOA矩阵法(SS-DOAM)。该算法利用M对阵元能够估计M-1个信号的二维波达方向,而且不论信号源是否相干都同样适用。由于只需要利用单快拍数据就可以估计出二维方向信息,所以它非常适合于瞬时信号、快速运动信号的二维DOA估计与跟踪。结合单快拍DOA矩阵法,又提出了一种新的多用户多径二维DOA估计方法。与基于时间平滑的方法不同,新方法不再局限于块衰落信道模型,并且大大降低了计算量。方法适用于双平行均匀线阵,无法用于其它阵列结构。研究了空时二维DOA估计技术,提出了基于均匀圆阵的空时矩阵束法和基于任意阵列的空时二维Unitary-ESPRIT算法。这两种算法最大的特点是都充分利用了信号在时空域的二维信息,都依赖特征值估计二维参数并实现自动配对,因此不可能发生角度兼并问题,同时它们都有很高的阵元利用率。但这两种算法解决问题的思路却截然不同,一种算法利用了矩阵束的思想,另一种算法利用了Unitary-ESPRIT的思想。空时矩阵束法估计二维参数时是通过广义特征值的幅度和相位进行求解和配对,而空时二维Unitary-ESPRIT法是利用特征值的实部和虚部进行求解和配对。相比之下后者计算量较大,但阵列结构灵活。提出了一种具有很强稳健性的自适应波束形成算法,该算法既适合于一维也适用于二维波束形成。它能对多径环境中的期望信号实现最大信干噪比(SINR)意义下的最优波束形成,而且只需要已知期望信号全部或部分入射路径的大致方向。这种方法既克服了盲算法无法辨识期望信号的缺点,又克服了非盲算法对系统误差敏感的缺点,是一种稳健性非常强的实用算法。更多还原

【Abstract】 This dissertation is devoted to a study of Two-dimensional Direction finding and Adaptive Beamfroming.The work finished in this paper is a part of a large scale army project of research undertaken by the lab the author works with.Direction of arrival(DOA)estimation and beamforming(BF),as kernel topics of array signal processing,have attracted considerable attentions.In this field,numerous achievements of theoretical research were attained in the past decades.Two-dimensional(2-D)direction finding (DF)and beamforming,compared with its one-dimensional(1-D)counterpart,utilize the spatial information of signals more fully,thus result in performance improvements in location accuracy and communication capacity,and have more practical values as well as even wider application fields.However in 2-D array processing,due to its complexity,there are many difficult problems to be solved such as angle ambiguity,aperture losing,DOA estimation of fast moving targets or short signals,DOA estimation in multiuser and multipath environments,robust adaptive BF etc.. This dissertation is focusing on 2-D direction finding and robust beamforming.Some novel or improved methods as solutions to the above mentioned problems are proposed.Angle ambiguity problem in DOA matrix methods and the like is analyzed first,and then some improved algorithms are proposed.A conclusion is drawn that angle ambiguity exists anywhere in 2-D DOA estimation methods based on rotational invariance and can never be avoided as long as the angle parameter is obtained directly by the use of steering vector,thus some auxiliary methods must be taken to solve the problem.Two algorithms of 2-D direction finding based on temporal smoothing are proposed for multiuser and multipath situations aiming at double parallel uniform linear array and uniform circular array respectively.The algorithms can make 2-D DOA estimations of all paths of non-Gaussian signals in Gaussian noise and can group DOAs according to their coherence,thus are suitable for mobile communication.Besides,the maximum number of estimated DOAs is much more than the number of sensors.The computational complexity of the temporal smoothing algorithms is high and the algorithms hold only for the channel model of block fading.A single-snapshot 2-D direction finding method based on double parallel uniform linear arrays,called SS-DOAM method,is brought forward for the first time in the field.DOAs of M-1 signals,no matter whether they are coherent or not,can be estimated using M pairs of elements. Requiring only one snapshot data,the algorithm is very suitable for instant signals and fast moving targets.Based on above SS-DOAM method,a new 2-D DOA estimation method for the situations of multiuser and multipath is proposed.Being different form the above proposed method based on temporal smoothing,this method is not constrained to block fading channel model and the computational complexity is also remarkably reduced.However this method can not be used for the arrays other than double parallel uniform linear array. In the aspect of space-time 2-D direction finding,two new algorithms,i.e.space-time matrix pencil algorithm for uniform circular array(UCA)and space-time 2-D Unitary-ESPRIT algorithm for arbitrary array,are proposed.The common features of them lie in that they utilize fully the spatial and temporal information of signals and they estimate 2-D parameters only by the use of eigenvalues,thus the angle ambiguity can be avoided completely,and the sensor utilization factor becomes higher.But the strategies of them are distinct.One of them uses matrix pencil technique and the other uses Unitary-ESPRIT.In Matrix pencil method 2-D parameters are obtained according to the magnitude and phase of generalized eigenvalues,while in Unitary-ESPRIT method 2-D parameters are obtained according to the imaginary and real parts of eigenvalues.By comparison,the latter one is more flexible in array configuration,however is higher in computational complexity.Finally,a robust adaptive beamforming method is provided in Chapter 7 which is suitable for both 1-D and 2-D beamforming and can implement optimal beamforming only using partial and rough DOA information of coherent paths of desired signal by the criterion of maximum SINR.The algorithm makes blind algorithms capable of identify expected signals and solved the problem of sensitivity to system errors of non-blind beamforming algorithm.Simulation results proved the feasibility and robustness of the proposed algorithm.更多还原