基于GAA模型的分布源直接定位方法
发布时间：2018-10-23 10:34
【摘要】：分布源定位是通信等领域的重要问题之一。实际应用中,点目标周围往往密集地分布着散射点,导致目标信号在空间上产生一定的角度扩展,传统基于点源假设的定位方法用于这类场合将导致目标定位性能下降。为此,需要针对具体分布式目标,建立合理的分布式目标参数化模型及后续定位方法。现有的分布源定位方法大都是基于观测量估计的两步定位方法,这类方法忽略了所有观测量需要对应于同一目标的约束条件,使得其在低信噪比时定位精度不高。针对这一问题,本文着眼于分布式目标的直接位置估计方法的学习和研究。本文主要内容如下:(1)分布式目标建模针对点源模型的局限,引入了角度扩展参数,建立分布式目标的模型;介绍相干和非相干分布源的概念,由此引出空间频率近似模型和两点源近似模型这两个简化的非相干分布源模型;针对实际通信场景中的非相干分布源,引入GAA(Gaussian Angle of Arrival)模型;假设目标源周围分布的散射点服从高斯分布,对其空间模型和时间模型进行了推导,得到了该信道模型下的阵列响应。(2)直接位置估计方法研究研究了目标源位置直接估计的基本理论。不同于传统两步定位框架中先进行中间参数的估计、再对目标位置进行估计的思路,直接定位方法直接基于原始接收数据对目标位置进行估计:通过数据的频域处理并基于最大似然准则,推导出相应的目标位置估计代价函数。针对多目标源场景,利用MVDR(Minimum Variance Distortionless Response)理论对信号进行滤波,提高了定位方法的分辨率。仿真结果表明:直接定位在低信噪比时具有较低的误差,MVDR方法具有良好的分辨率。(3)分布源直接定位方法研究将点源目标位置直接估计方法推广至分布源模型中,利用GAA模型对接收信号进行建模,基于最大似然准则得到分布源位置和散射半径的估计。在此基础上,研究几种次优的估计——牺牲少量定位精度,获得更低计算复杂度。在基于代价函数进行最优化求解时,利用梯度下降法对最优目标位置值进行迭代搜索。几种方法的仿真结果表明:最大似然估计具有最优精度,最小二乘估计兼具高定位精度和较低的计算量。
[Abstract]:Distributed source location is one of the important problems in communication and other fields. In practical applications, scattered points are often distributed around the point targets, which leads to a certain angle expansion of the target signal in space. The traditional localization method based on point source assumption will lead to the degradation of target localization performance. Therefore, it is necessary to establish a reasonable parameterized model of distributed target and its subsequent localization method. Most of the existing distributed source localization methods are two-step localization methods based on the estimation of observations. This method ignores all the constraints corresponding to the same target and makes the localization accuracy low when the SNR is low. To solve this problem, this paper focuses on the study and research of direct position estimation for distributed targets. The main contents of this paper are as follows: (1) aiming at the limitation of point source model, distributed object modeling introduces angle extension parameters to establish distributed target model, and introduces the concepts of coherent and incoherent distributed sources. Two simplified incoherent distributed source models, namely spatial frequency approximation model and two-point source approximation model, are derived, and GAA (Gaussian Angle of Arrival) model is introduced for incoherent distribution sources in actual communication scenarios. Assuming that the scattering points around the target source are distributed from Gao Si, the spatial model and the time model are derived. The array response of the channel model is obtained. (2) the direct position estimation method is used to study the basic theory of direct target source location estimation. Different from the traditional two-step localization framework, the intermediate parameters are estimated first, and then the target position is estimated. The direct location method estimates the target position directly based on the original received data. By processing the data in frequency domain and based on the maximum likelihood criterion, the corresponding target location estimation cost function is derived. The MVDR (Minimum Variance Distortionless Response) theory is used to filter the signal for multi-target source scene, which improves the resolution of the localization method. The simulation results show that direct location has lower error when SNR is low, and MVDR method has good resolution. (3) Direct location method of distributed source is extended to the distributed source model. The GAA model is used to model the received signal, and the location and scattering radius of the distributed source are estimated based on the maximum likelihood criterion. On this basis, several sub-optimal estimators are studied, which sacrifice a small amount of positioning accuracy to obtain lower computational complexity. When solving optimization based on cost function, the gradient descent method is used to iteratively search the optimal target position. The simulation results of several methods show that the maximum likelihood estimation has the best accuracy, and the least square estimation has both high positioning accuracy and low computational complexity.
【学位授予单位】：电子科技大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TN92
本文编号：2288979
[Abstract]:Distributed source location is one of the important problems in communication and other fields. In practical applications, scattered points are often distributed around the point targets, which leads to a certain angle expansion of the target signal in space. The traditional localization method based on point source assumption will lead to the degradation of target localization performance. Therefore, it is necessary to establish a reasonable parameterized model of distributed target and its subsequent localization method. Most of the existing distributed source localization methods are two-step localization methods based on the estimation of observations. This method ignores all the constraints corresponding to the same target and makes the localization accuracy low when the SNR is low. To solve this problem, this paper focuses on the study and research of direct position estimation for distributed targets. The main contents of this paper are as follows: (1) aiming at the limitation of point source model, distributed object modeling introduces angle extension parameters to establish distributed target model, and introduces the concepts of coherent and incoherent distributed sources. Two simplified incoherent distributed source models, namely spatial frequency approximation model and two-point source approximation model, are derived, and GAA (Gaussian Angle of Arrival) model is introduced for incoherent distribution sources in actual communication scenarios. Assuming that the scattering points around the target source are distributed from Gao Si, the spatial model and the time model are derived. The array response of the channel model is obtained. (2) the direct position estimation method is used to study the basic theory of direct target source location estimation. Different from the traditional two-step localization framework, the intermediate parameters are estimated first, and then the target position is estimated. The direct location method estimates the target position directly based on the original received data. By processing the data in frequency domain and based on the maximum likelihood criterion, the corresponding target location estimation cost function is derived. The MVDR (Minimum Variance Distortionless Response) theory is used to filter the signal for multi-target source scene, which improves the resolution of the localization method. The simulation results show that direct location has lower error when SNR is low, and MVDR method has good resolution. (3) Direct location method of distributed source is extended to the distributed source model. The GAA model is used to model the received signal, and the location and scattering radius of the distributed source are estimated based on the maximum likelihood criterion. On this basis, several sub-optimal estimators are studied, which sacrifice a small amount of positioning accuracy to obtain lower computational complexity. When solving optimization based on cost function, the gradient descent method is used to iteratively search the optimal target position. The simulation results of several methods show that the maximum likelihood estimation has the best accuracy, and the least square estimation has both high positioning accuracy and low computational complexity.
【学位授予单位】：电子科技大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TN92
【参考文献】
相关期刊论文 前5条
1 陈鸾;张海剑;孙洪;;多重目标直接定位的子空间分解压缩感知算法[J];信号处理;2015年10期
2 王云龙;吴瑛;;联合时延与多普勒频率的直接定位改进算法[J];西安交通大学学报;2015年04期
3 万群,杨万麟;一种相干信号源分布式目标波达方向估计方法[J];系统工程与电子技术;2001年03期
4 万群,杨万麟;一种分布式目标波达方向估计方法[J];通信学报;2001年02期
5 万群,杨万麟;相干分布式目标一维波达方向搜索迭代估计方法[J];电子科技大学学报;2000年06期
，本文编号：2288979
本文链接：https://www.wllwen.com/kejilunwen/xinxigongchenglunwen/2288979.html
