【作者】 赵大勇；

【导师】 杨莘元；

【作者基本信息】 哈尔滨工程大学 ， 通信与信息系统， 2011， 博士

【摘要】 阵列信号处理中的DOA(Direction Of Arrival)估计技术有着很大的应用前景,无论是雷达、通信、声纳,还是地震勘探、射电天文等领域对它都非常青睐。其中智能天线空分多址、声源定位探测以及无源阵列雷达等学科已经把它列为了一项重要技术。各国的专家学者在过去的几十年里陆续提出了很多性能良好的方法,可谓硕果累累。然而由于一些难点没有得到解决,许多的方法还不能真正的用到实际系统上。论文介绍了阵列信号处理的基本模型以及一些理论方法,并且对已有的一些DOA估计算法的特点进行了分析,针对它们的不足之处提出了新的改进算法。在实际应用中,空间存在大量波达方向随时间变化的信号源,因此针对运动目标的动态DOA估计也就成为波达方向估计中的重要研究课题,最大似然估计(MLE: Maximum Likelihood Estimation)在DOA处理中具有很好的性能,然而它需要进行多维参数的搜索,所以运算起来非常耗时。针对信号方向时刻变化这一情况,提出了一种基于粒子群优化技术的跟踪方法,这种方法能够自动的对目标进行跟踪,并在很小的一个区域内锁定目标,回避了子空间类方法需要重复分解协方差矩阵的过程。这样便在很大的程度上使计算量得到了降低。而且算法有着较好的跟踪精度和实时性能,还能处理相干信号。一般的DOA估计方法需要假设信号以及背景噪声为高斯白噪声,这样便可以利用二阶协方差矩阵的方法很方便的估计出信号方向。但是实际生活里的很多信号以及噪声往往不是服从高斯分布,比方说通信线路上的瞬间干扰、海洋环境噪声以及大气放电产生的噪声,还有许多人为噪声等等,它们都有着明显的尖峰,如果仍然采用二阶协方差矩阵的方法来处理,那么肯定得不到理想的估计,幸亏有一种称为α稳定分布的信号模型可以对以上的信号进行描述。论文对α稳定分布进行了介绍,然后分析了描述噪声的分数低阶矩这一模型。针对动态DOA估计这一问题,基于粒子群优化技术提出了一种新的锁定跟踪方法。它不必重复的分解分数低阶矩矩阵,使多维搜索的计算量得到了有效的降低,而且跟踪性能也很好。由于通常的二维DOA估计都需要较多的阵元,然而利用率都不高。在最小冗余线阵的基础上提出了一种新的阵列模型,模型设置了三条平行阵列,它的阵元冗余度很低,并且结合传播算子方法实现了信号二维DOA估计,避免了矩阵特征分解以及谱峰搜索过程。由于采用了最小冗余线阵,所以阵列孔径得到了增大,如此便使传播算子方法在低信噪比下也获得了较好的性能。另外为了能够更好的在冲击噪声背景下对信号DOA进行估计,论文结合分数低阶矩矩阵以及最小冗余线阵提出了一种新颖的人工蜂群算法。算法使阵列孔径得到了扩展,它的分辨力在冲击噪声背景下也有较好的表现,而且使用的阵元数目可以少于信号数目。此外在L形阵列的基础上提出了新的二维DOA估计方法,它利用了参考阵列的旋转不变性,通过纵向和横向两重的扩展构造了虚拟阵列,而且该虚拟阵列也有着旋转不变性,与此同时它的四阶累积量矩阵不存在数据冗余。新算法估计信号子空间时只需完成两个四阶累积量矩阵的计算,再结合ESPRIT方法即可得到信号方向。新方法具有精度高、计算量低的优点,论文从理论和实验两个方面证明了算法的性能。在宽带DOA估计算法中,聚焦矩阵的选取非常重要,它直接影响着估计性能。论文提出了一种构造聚焦矩阵的新方法,使得信噪比较低时也能够较好的完成DOA估计,且聚焦前后没有损失信噪比。实验证明了该方法比传统的双边相关变换(TCT:Two-sided Correlation Transformation)算法的估计精度更高。更多还原

【Abstract】 Direction-Of-Arrival(DOA) estimation techniques in array signal processing with array antenna have wide applications in a variety of fields ranging from radar, communication, sonar, seismology to radio astronomy. Especially, they become a key technique in the passive detection of array radar, SDMA of smart antenna system and detection sound source. Since early 1980s, high resolution DOA estimation techniques have received considerable attention and a lot of significant processes have been achieved in this field. However, there are still important and urgent problems that have not been solved perfectly. The basic model and theory of array signal processing are introduced in this dissertation firstly, then, the properties and disadvantages of the existed DOA estimation are analyzed. Furthermore, the new algorithms are proposed and discussed.In actual situations, there are many moving sources, so dynamic DOA estimation becomes the important research topic, MLE has the excellence performance as a DOA method, it is a no linearity and multidimensional estimation, it take a long time to do, A new method to estimate direction-of-arrival (DOA) of moving sources is proposed. Making use of maximum likelihood algorithm, this method can avoid the decompositions of the covariance matrix which should be repeated in the methods based on subspace tracking. In order to solve the problem of the huge computation cost in maximum likelihood algorithm, the particle swarm algorithm was considered and improved. So the aims can be tracked and estimated in a very little space in which the maximum is searched for. In this way the searching place was reduced greatly and the swarm intelligence was used in searching, so the cost can be reduced mostly. Simulation results show that the DOA estimation based on the improved particle swarm algorithm has the ability to track coherent sources and performs better than the methods based on subspace tracking in the aspect of tracking precision with the ability of being real-time.The traditional algorithms always run into the supposition that the signal and the noise obey the Gaussian distribution, and obtain better results by using more than second-order statistics. In practical applications, much random signal and noise encountered is not Gaussian distribution, such as atmospheric and lightning noise, the instantaneous peak on communications line and a variety of man-made noise, in which these are many significant peaks and the traditional second-order statistics-based methods of treatment should not be satisfied. There is a very important statistical signal model known as the Alpha stable distribution, which can describe the above-mentioned noise. Therefore the text briefly introduced the stable distribution and the fractional lower order moment. This paper provides a new improvement on particle swarm optimization algorithm basing on the idea of locking and tracking, and study a new method based on the maximum likelihood algorithm for dynamic direction-of-arrival (DOA) estimation in impulsive noise environments. This method can avoid the decompositions of the fractional lower order moments matrix which should be repeated in the methods based on subspace tracking, and perform better than the methods based on subspace tracking in the aspect of tracking precision. In addition the cost of the multidimensional search can be reduced mostly.In order to overcome the low sensor utilization rate problem existing in most two-dimensional DOA estimation algorithms, a new array model with low array redundancy is proposed in this paper. Therefore, the application of MRLA is extended to 2-D DOA estimation. Simultaneously, two-dimensional spectral peak searching and Eigen-decomposition of large matrix is avoided by using propagator method. The computational complexity is greatly reduced. The larger array effective aperture is obtained by using MRLA, So the performance of DOA estimation in the poor environment with low SNR is obviously improved. Simulation results showed the superiority of this proposed method in precision. Based on virtual multi-element uniform linear array and reconstructed fractional lower order covariance matrix, a novel maximum likelihood (ML) algorithm is proposed. The proposed algorithm utilized few virtual elements and expanded the number of effective aperture array, while significantly improving the performance of the original ML algorithms, In order to fit the proposed direction finding algorithm based on the minimum redundant array and fractional lower order covariance matrix, a bee colony algorithm is applied to objective function of direction finding. Monte-Carlo simulations have proved that the proposed method has some good performance such as high resolution in the presence of impulse noise and the capability of using a small number of elements to find more signal sources. A new method for estimating two-dimensional direction-of-arrival based on special linear array was presented. The virtual array’s rotational invariance can be got by the rotational invariance of the reference arrays, by using the method of two directions’expansion, the array’s expanding ability can be increased, and it can also eliminate the redundant data of the forth-order cumulant matrix. By using this new method, the signal’s subspace estimation can be obtained by only two forth-order cumulant matrix, and the estimation of the signal’s DOA can be achieved by the 2-D ESPRIT method. The theoretical analysis and simulated results show that this method is characterized by low computation cost, well expanding ability、high precision and good practicability.In the wide-band direction-finding algorithms based on the signal subspace approach, the focusing matrix has an important effect on the performance of the estimation. This paper proposed a new method of constructing focusing matrix and the signal to noise ratio of the array before and after focusing was equal without any loss. New approach can distinguish the two targets which are close to each other even break the restriction of Rayleigh resolution limit and has higher accuracy compared to TCT algorithm while the signal to noise ratio is very low. Extensive simulation results demonstrate that the algorithm has good performance.更多还原