【作者】 李福昌；

【导师】 赵春晖；

【作者基本信息】 哈尔滨工程大学 ， 信号与信息处理， 2005， 博士

【摘要】 波达方向(DOA)估计在移动和卫星通信系统、信息战、雷达、被动声纳、地震学、射电天文学、导航、视频会议、时间序列分析、谱估计等方面有着广泛的应用,引起了人们极大的研究兴趣。但已有的测向算法,大多数是基于被测信号为窄带信号这一前提假设的,也即信号的带宽远小于中心频率,信号复包络在阵列的各个传感器上被视为是无差别的,由于信号到达各个传感器上的时间延迟不同而产生相位上的差异,通过时域处理就可以对波达方向进行估计。 随着通信技术的发展,跳频信号、扩频信号、线性调频信号等宽带信号在通信系统中的应用越来越多;另外,在自然界中还有许多信号本质上就属于宽带信号,如声音信号、地震波等。因此,研究宽带信号测向就显得越来越重要。 本文针对宽带信号的特点,利用宽带信号时宽—带宽乘积大的特点,分别研究了宽带最大似然测向算法、基于波达方向预估计的宽带子空间测向算法、波达方向信息完全未知情况下的宽带子空间测向算法和宽带波束形成器,从理论上分析了它们的性能,用仿真实验对其波达方向估计性能进行验证。本文的主要研究工作和取得的主要成果如下: (1)在高斯噪声假定下,基于最大似然的宽带测向算法是最优的,但代价函数的优化是个很难处理的问题。本文提出了把交替投影算法和EM算法结合起来对宽带确定性最大似然测向算法代价函数进行优化,充分利用了交替投影算法收敛率快的特性,同时消除了EM算法的初值敏感性。 (2)提出了应用遗传算法对宽带确定性最大似然测向算法代价函数和宽带加权子空间拟合测向算法代价函数进行优化,得出了波达方向估计结果与遗传算法运行参数之间的关系,为宽带测向的实际工程应用提供了依据。 (3)本文对传统的旋转信号子空间宽带测向算法进行了修正,既减少了聚焦矩阵的运算量,又提高了波达方向估计精度和分辨能力。 (4)通常的基于阵列流行变换的宽带测向算法,在求聚焦矩阵时需要对更多还原

【Abstract】 DOA (direction of arrival) estimation has wide applications in mobile and satellite communication systems, information warfare, radar, passive sonar, seismology, astronomy, navigation, video conferencing, time series analysis and spectrum estimation, and has received much attention in recent years. But the current literature about direction finding algorithms pays more attention to narrow-band signal, that is to say bandwidth of signal is far smaller than center frequency of signal and so complex envelop of signal keeps invariant in the sensors of the array. The signal arrives on different sensors at different time delays and so phase difference is caused, then DOA is estimated in time domain.With the development of communication technologies, wide-band signals are used more and more widely in communication systems. In the other hand, there are many wide-band signals in nature, such as seismology wave, acoustic signal. So the research on wide-band direction finding becomes more and more important.Based on the characteristic of wide-band signal, this dissertation makes full use of the large time-bandwidth product and proposes wide-band maximum likelihood direction finding algorithms, wide-band sub-space direction finding algorithms based on DOA pre-estimation, wide-band direction finding algorithms under the condition of complete not knowing DOA information and wide-band beamformer, analyzing their performance theoretically and verifying it through simulation. The main research work and contributions of the dissertation are as follows:(1) Under the assumption of Gaussian noise, direction finding algorithms based on maximum likelihood are optimal, but the optimization of cost function is an intractable problem. The dissertation combines alternating projection algorithm and EM algorithm to optimize the wide-band deterministic maximum likelihood direction finding algorithm cost function, making full use of fast convergence rateof alternating projection algorithm and eliminating the sensitivity of EM algorithm to initial value.(2) The dissertation applies genetic algorithm to optimize the cost function of wide-band deterministic maximum likelihood direction finding and wide-band weighted sub-space fitting direction finding, and attains the relation between DOA estimation results and operation coefficients of genetic algorithm, which supplies basis for the engineering application of wide-band direction finding.(3) The dissertation modifies traditional RSS wide-band direction finding algorithm, not only attaining lower complexity of computation, but also improving estimation precision and resolution ability.(4) In traditional wide-band direction finding algorithms based on array manifold transform, the computation of focusing matrix needs DOA pre-estimation, but there is often bias in pre-estimation, so it can lead to uncertain effect on the final estimation result. The dissertation proposes wide-band direction finding algorithm based on noiseless covariance matrix transform, because the computation of transform matrix uses sensors data directly, so there is no effect of pre-estimation bias on the final result.(5) The dissertation proposes wide-band direction finding algorithm based on pseudo-covariance matrix eigen-decompositioin. A pseudo-data matrix and then a pseudo-covariance are constructed using the signal eigen-vectors of eigen-decomposition of narrow-band covariance matrix. After the eigen-decomposition of the pseudo-covariance matrix, noise sub-space is got and sub-space method is utilized to get DOA estimation.(6) The dissertation proposes the improved version of wide-band steered minimum variance beamformer. Simulation results based on ULA mode demonstrates that, compared with the original algorithm, the performance of improved algorithm improves in estimation precision and resolution ability.更多还原