基于原子范数最小化的限定区域波达方向快速估计
Fast constrained area direction-of-arrival estimation based on atomic norm minimization
-
摘要: 受限于有界引理的应用条件, 基于原子范数最小化的波达方向估计方法通常只能在全区域进行角度搜索, 带来大量的运算冗余。针对这一问题, 通过对对偶变量进行带通滤波和降采样, 使得相应的对偶三角多项式不等式约束可以使用有界引理转化为半正定约束, 从而将对偶问题转化为半正定规划问题, 进而得到了基于原子范数最小化的限定区域波达方向估计方法。为了提升其运算效率, 还提出了一种适用于该算法的快速求解方法, 通过原对偶内点法提升了运算速度。仿真结果表明, 相比于原始原子范数最小化算法, 所提限定区域的原子范数最小化算法的运算时间更短; 所提快速求解算法在估计精度与原算法一致的条件下运算速度进一步提升。Abstract: Due to the constraints of the bounded lemma, direction-of-arrival (DOA) estimation methods based on atomic norm minimization (ANM) usually necessitates exhaustive search over the entire angular domain, resulting in substantial computational redundancy. To address this challenge, this paper proposes a dual-variable processing framework involving bandpass filtering and downsampling. This approach enables the conversion of dual trigonometric polynomial inequality constraints into semidefinite constraints through the bounded lemma, thereby transforming the dual problem into a semidefinite programming (SDP) formulation. Consequently, an ANM-based DOA estimation method with angular sector constraints is developed. To enhance computational efficiency, this paper further propose a fast algorithm that leverages primal-dual interior-point methods to accelerate computation. Simulation results demonstrate that, compared to the original ANM algorithm, the proposed constrained area ANM algorithm significantly reduces computational time. Additionally, the proposed fast algorithm further increases computation speed while maintaining the same estimation accuracy as the original algorithm.