基于压缩感知理论的DOA估计与跟踪算法研究

发布时间：2018-11-27 19:21

【摘要】：随着阵列信号DOA估计与跟踪技术在雷达探测和声呐定位等实际场景中的应用愈来愈广泛和深入,人们对DOA估计与跟踪技术的要求也日益提高。近年来,随着压缩感知理论的发展,其对快拍数要求低和天然的解相干性等鲜明的优点促进了阵列信号DOA估计技术的不断深入研究;并且,随着动态压缩感知等理论的形成和不断完善,基于压缩感知理论的DOA跟踪技术也取得了一定的研究成果。本文在这一背景下对基于压缩感知理论的DOA估计和跟踪算法进行了相关研究,主要研究内容归纳如下:1)通过对平滑l0范数的研究,采用改进后性能更加优异的平滑函数来逼近l0范数,提出一种基于改进平滑l0范数的DOA估计算法,该算法容易实现且精度较高,在单快拍条件下就能对DOA进行较好的估计,且相比OMP算法和平滑l0范数原始算法具有更好的性能。2)通过对改进平滑l0范数采用新的加权方式进行处理,提出一种基于加权平滑l0范数的DOA估计算法,该算法同样容易实现,在单快拍条件下就能实现较高精度的DOA估计,且相比基于改进平滑l0范数的DOA估计算法具有更高的估计精度。3)实际应用中的DOA估计技术通常为MMV模型,本文将基于加权平滑l0范数的DOA估计算法推广到MMV模型下,提出一种基于多快拍加权平滑l0范数的DOA估计算法,该算法在较低快拍数条件下就可实现DOA的高精度估计。4)针对运动目标信号源的DOA跟踪问题,将动态压缩感知理论的处理方法应用到动态DOA这一时变稀疏信号中,并建立动态DOA稀疏概率模型以获得加权l,范数的权值,最终通过对线性加权l1范数的最小化,提出一种单快拍情况下的动态压缩感知DOA跟踪算法,该算法可实现较高精度的DOA跟踪,在一定信噪比条件下具有比PASTd算法和粒子滤波算法更好的DOA跟踪性能。5)为提升动态压缩感知DOA跟踪算法对噪声的抗干扰能力,将该算法推广到MMV模型下,同时对接收信号进行奇异值分解处理以降低计算量,最终提出一种多快拍动态压缩感知DOA跟踪算法,该算法可以在快拍数较少且信噪比较低的情况下实现较高精度的DOA跟踪。
[Abstract]:With the increasing application of array signal DOA estimation and tracking technology in radar detection and sonar localization, the requirements of DOA estimation and tracking technology are increasing. In recent years, with the development of compression sensing theory, its advantages such as low requirement of rapid-beat number and natural desiccation have promoted the further research of array signal DOA estimation technology. Moreover, with the formation and improvement of the theory of dynamic compression sensing, the DOA tracking technology based on the theory of compressed sensing has also achieved some research results. In this context, the DOA estimation and tracking algorithm based on compressed perception theory is studied in this paper. The main research contents are summarized as follows: 1) through the research of smoothing l0 norm, The improved smoothing function is used to approximate the l0 norm, and a DOA estimation algorithm based on the improved smoothing l0 norm is proposed. The algorithm is easy to implement and has high accuracy. The DOA can be estimated better under the condition of single shot. Compared with the OMP algorithm and the original smoothing l0 norm algorithm, it has better performance. 2) A new DOA estimation algorithm based on the weighted smoothing l0 norm is proposed by using a new weighting method for the improved smoothing l0 norm. This algorithm is also easy to implement, and can achieve high precision DOA estimation under the condition of single beat. Compared with the DOA estimation algorithm based on improved smoothing l0 norm, it has higher estimation accuracy. 3) the DOA estimation technique in practical application is usually MMV model. In this paper, the DOA estimation algorithm based on weighted smoothing l0 norm is extended to MMV model. A new DOA estimation algorithm based on multi-beat weighted smoothing l0 norm is proposed. The algorithm can realize the high precision estimation of DOA under the condition of lower beat number. 4) aiming at the DOA tracking problem of moving target signal source, a new algorithm is proposed. The processing method of dynamic compression sensing theory is applied to the transient sparse signal of dynamic DOA, and the sparse probability model of dynamic DOA is established to obtain the weight of weighted L and norm. Finally, the linear weighted L 1 norm is minimized. This paper presents a dynamic compression sensing DOA tracking algorithm in the case of single racket, which can achieve high precision DOA tracking. Under certain SNR conditions, the DOA tracking performance is better than that of PASTd algorithm and particle filter algorithm. 5) in order to improve the anti-jamming ability of dynamic compression sensing DOA tracking algorithm to noise, the algorithm is extended to MMV model. At the same time, the received signal is processed by singular value decomposition to reduce the computational complexity. Finally, a multi-beat dynamic compression sensing DOA tracking algorithm is proposed, which can achieve high precision DOA tracking under the condition of fewer beats and lower signal-to-noise ratio (SNR).
【学位授予单位】：吉林大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TN911.23

【相似文献】