改进的变分稀疏贝叶斯学习离格DOA估计方法

Improved variational sparse Bayesian learning off-grid DOA estimation method

为提高阵列信号处理运算速率,改善其方位估计性能,提出了一种改进变分稀疏贝叶斯学习离格波达方向(direction-of-arrival, DOA)估计方法。该方法利用实值变换,将向量化后的接收信号协方差矩阵转化到实数域,结合变分稀疏贝叶斯学习和网格演化的思想,在迭代过程中使网格从初始的均匀网格自适应地演化为非均匀网格,通过网格更新和网格裂变交替迭代使演化后的网格点逐渐逼近真实信源方位。仿真结果表明,改进方法与传统压缩感知类方法相比,减小了运算量,提高了运算速率,且具有更高的方位估计精度和方位分辨能力,在少快拍和低信噪比情况下,改进方法性能提升的优势更明显。湖上试验数据处理结果进一步验证了该方法的有效性和工程实用性。

Here,to improve processing speed and direction-of-arrival(DOA)estimation performance of array signals,an improved variational sparse Bayesian learning off-grid DOA estimation method was proposed.This method could utilize real value transformation to transform covariance matrix of vectorized receival signals in complex domain into real domain.Ideas of variational sparse Bayesian learning and grid evolution were combined to make a grid adaptively evolute from an initial uniform one to a non-uniform one in iteration process.Though grid update and grid fission alternating iterations,evolved grid points could gradually approach DOA of actual signal source.Simulation results showed that compared with traditional compressed sensing methods,the proposed method can reduce computational amount,improve computational speed,and have higher DOA estimation accuracy and DOA resolution;in the case of fewer snapshots and low signal-to-noise ratio,these advantages become more obvious;data processing results of on-lake tests further verify the effectiveness and engineering practicality of the proposed method.