Accelerating Parallel Jacobi Method for Matrix Eigenvalue Computation in DOA Estimation Algorithm
Accelerating Parallel Jacobi Method for Matrix Eigenvalue Computation in DOA Estimation Algorithm
复制标题
DOA估计算法中矩阵特征值计算的加速并行雅可比法
DOI：
10.1109/tvt.2020.2984705
复制
发表时间：
2020-06-01
影响因子：
6.8
通讯作者：
Liu, Ying
中科院分区：
文献类型：
--
作者：
Shi, Zhiguo;He, Qianwen;Liu, Ying
许多算法都需要计算矩阵的特征值。具体地说，它是基于子空间的波达方向（DOA）估计算法中的关键技术，例如，多信号分类（MUSIC）。因此，特征值的计算直接影响DOA估计方法的实时实现。然而，经典的Jacobi方法是耗时的。在文献中，已经采用并行实现来加速特征值的计算。在本文中，我们建议进一步减少这种并行方法的执行时间。特别是，所提出的方法的每个并行单元每次迭代使用一个坐标旋转数字计算机（CORDIC）周期，而更多的是由传统的同行所需要的，这样的MUSIC算法的特征值分解可以加速。最后，在FPGA平台上实现了该方法。实验结果表明，该方法具有更高的计算效率。
The calculation of eigenvalues of a matrix is required by many algorithms. Specifically, it is the key technique in subspace-based direction of arrival (DOA) estimation algorithms, e.g., multiple signal classification (MUSIC). The calculation of the eigenvalues therefore directly affects the real-time implementation of DOA estimation approaches. However, the classical Jacobi methods are time-consuming. In literature, a parallel implementation has been adopted to accelerate the calculation of eigenvalues. In this paper, we propose to further decrease the execution time of this parallel method. In particular, each parallel unit of the proposed method uses one coordinate rotation digital computer (CORDIC) period per iteration, while more are required by the traditional counterparts, such that the eigenvalue decomposition of the MUSIC algorithm can be accelerated. In addition, the proposed method is implemented in an FPGA platform. The experimental results show that the proposed method is more computationally efficient.