National Key Laboratory of Mechatronic Engineering and Control, Beijing Institute of Technology, Beijing 100081, China.
Sensors (Basel). 2018 Nov 28;18(12):4180. doi: 10.3390/s18124180.
Angle estimation methods in two-dimensional co-prime planar arrays have been discussed mainly based on peak searching and sparse recovery. Peak searching methods suffer from heavy computational complexity and sparse recovery methods face some problems in selecting the regularization parameters. In this paper, we propose an improved trilinear model-based method for angle estimation for co-prime planar arrays in the view of trilinear decomposition, namely parallel factor analysis. Due to the principle of trilinear decomposition, our method does not require peak searching and can conduct auto-pairing easily, which can reduce the computational loads and avoid parameter selection problems. Furthermore, we exploit the virtual array concept of the whole co-prime planar array through the cross-correlation matrix obtained from the received signal data and present a matrix reconstruction method using the Khatri⁻Rao product to tackle the matrix rank deficiency problem in the virtual array condition. The simulation results show that our proposed method can not only achieve high estimation accuracy with low complexity compared to other similar approaches, but also utilize limited sensor number to implement the angle estimation tasks.
本文主要讨论了基于峰值搜索和稀疏恢复的二维互质平面阵列的角度估计方法。峰值搜索方法计算复杂度高，稀疏恢复方法在选择正则化参数时面临一些问题。在本文中，我们提出了一种改进的基于三线性模型的互质平面阵列角度估计方法，从三线性分解的角度来看，即平行因子分析。由于三线性分解的原理，我们的方法不需要峰值搜索，并且可以轻松进行自动配对，从而降低了计算负载并避免了参数选择问题。此外，我们通过从接收信号数据获得的互相关矩阵利用整个互质平面阵列的虚拟阵列概念，并提出了一种使用 Khatri⁻Rao 积的矩阵重建方法来解决虚拟阵列条件下的矩阵秩不足问题。仿真结果表明，与其他类似方法相比，我们提出的方法不仅具有较低的复杂度和较高的估计精度，而且还可以利用有限的传感器数量来实现角度估计任务。
