贝叶斯压缩感知在DOA估计中的应用研究

发布时间：2018-06-06 14:24

本文选题：波达角度估计 + 贝叶斯压缩感知　；参考：《哈尔滨工业大学》2014年硕士论文

【摘要】：作为阵列信号处理理论的重要研究方向之一，信号波达角度（Direction ofArrival，DOA）估计广泛应用于军事及国民经济各个领域。经典DOA估计算法需要信号源数目的先验知识和多快拍数的采样数据，，这在实际应用环境中较难实现，且当辐射信号源具有相关性时，经典算法的估计性能急剧下降甚至完全失效。贝叶斯压缩感知（Bayesian Compressive Sensing, BCS）为阵列信号DOA估计提供了一个新的解决思路，该理论突破了Nyquist采样定理的瓶颈，以基于统计模型的思想实现信号重构。本文主要研究基于BCS的DOA估计，以克服传统经典算法的不足。首先，结合窄带远场阵列信号DOA估计模型，分别构建了两类基于BCS的DOA估计：一是实现实时估计的单快拍模型；二是获取高分辨的多快拍模型，在此基础上为了降低计算复杂度研究讨论了对应的快速算法。由于噪声参数的辨识性作用显著影响算法性能，为了改善这一缺陷，提高算法的鲁棒性，进一步研究了修正的贝叶斯压缩感知理论。基于以上估计模型，分别从信噪比、信号源角度间隔、阵元数、快拍数、相干信号源场景和计算复杂度六个方面仿真分析比较了经典MUSIC算法、正交匹配追踪（Orthogonal Matching Pursuit，OMP）算法以及BCS算法的估计性能。最后，从信号源稀疏表示和超参数先验分布两个方面对BCS算法进行改进，分别得到两类改进算法：（1）基于等正弦网格划分的改进算法，理论证明了同传统等角度划分方式相比，基于等正弦网格划分对应的阵列流形矩阵具有更优的不一致性(Mutual Incoherence Property, MIP)。（2）基于Laplace先验的BCS-DOA估计，分别构建了基于Laplace先验的单快拍和多快拍条件下的DOA估计模型。在此基础上，从仿真分析的角度验证了两种改进算法的DOA估计性能优于原始BCS算法，表明了两类改进算法的可行性。
[Abstract]:As one of the important research directions of array signal processing theory, signal DOA estimation is widely used in various fields of military and national economy. The classical DOA estimation algorithm requires prior knowledge of the number of signal sources and sampling data of multiple beats, which is difficult to implement in practical applications, and when the radiation signal source is correlated, The estimation performance of the classical algorithm drops sharply and even fails completely. Bayesian Compressive Sensing, BCS) provides a new solution for array signal DOA estimation. The theory breaks through the bottleneck of Nyquist sampling theorem and realizes signal reconstruction based on statistical model. In this paper, DOA estimation based on BCS is studied to overcome the shortcomings of traditional classical algorithms. Firstly, two kinds of DOA estimation models based on BCS are constructed by combining with the DOA estimation model of narrowband far-field array signals: one is a single-shot model to realize real-time estimation, the other is to obtain a high-resolution multi-shot model. In order to reduce the computational complexity, the corresponding fast algorithm is discussed. In order to improve the robustness and improve the robustness of the algorithm, the modified Bayesian compression sensing theory is further studied in order to improve the performance of the algorithm because of the effect of the identification of the noise parameters on the performance of the algorithm. Based on the above estimation model, the classical MUSIC algorithm is simulated and compared from six aspects, such as signal-to-noise ratio (SNR), angle interval of signal source, array element number, rapid-beat number, scene of coherent signal source and computational complexity. The estimation performance of orthogonal Matching pursuit (OMP) algorithm and BCS algorithm. Finally, the BCS algorithm is improved from two aspects of signal source sparse representation and super-parameter prior distribution, and two improved algorithms, one of which is based on equal-sinusoidal mesh generation, are obtained, respectively. The theoretical proof is compared with the traditional equal-angle partition method. Based on the BCS-DOA estimation of Laplace priori, the DOA estimation models based on Laplace priori single rapid-beat and multi-rapid-beat are constructed, respectively, based on the array manifold matrix corresponding to the equal-sinusoidal mesh division with more inconsistency and mutual Incoherence property, and based on the BCS-DOA estimator of Laplace priori, respectively. On this basis, from the point of view of simulation analysis, the DOA estimation performance of the two improved algorithms is better than that of the original BCS algorithm, which shows the feasibility of the two improved algorithms.
【学位授予单位】：哈尔滨工业大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TN911.23

【参考文献】