一种利用稀疏贝叶斯学习的低复杂度DOA估计算法

A Low Complexity DOA Estimation Algorithm Utilizing Sparse Bayesian Learning

针对有限快拍数下传统利用子空间的多信号分类(MUSIC)算法估计信号波达方向(DOA)性能差,以及采用稀疏贝叶斯学习算法估计信号DOA时复杂度高的问题,提出一种联合利用稀疏贝叶斯学习和消息传递的高精度低复杂度DOA估计算法。该算法首先对信号进行实值化处理,其次把信号因子化为标量,最后使用消息传递策略避开稀疏贝叶斯学习算法每次迭代时的大矩阵求逆,降低计算复杂度。通过仿真对所提出的算法和经典算法进行对比验证,发现该算法在有限快拍数下不仅可以适用稀疏阵列的DOA欠定估计,而且在CPU运算时间方面与空间平滑MUSIC算法相当,且估计精度更高。

To solve the problem that the traditional subspace-based multi-signal classification(MUSIC)algorithm has poor performance in estimating the signal direction of arrival(DOA)and the sparse Bayesian learning algorithm has high complexity during limited snapshots,a high-precision and lowcomplexity DOA estimation algorithm utilizing sparse Bayesian learning and message passing is pro⁃posed.In this algorithm,the signal is real-valued first,and then the signal is factorized into a scalar.Fi⁃nally,the message passing strategy is used to avoid the large matrix inversion in each iteration of the sparse Bayesian learning algorithm,thus greatly reducing the computational complexity.The proposed algorithm is compared and validated with classical algorithms through simulation,and it has been found that the algorithm can not only be applied to underdetermined estimation of the DOA of sparse arrays under a limited number of snapshots,but also has higher estimation accuracy while maintaining the same CPU computation time as the spatial smoothing MUSIC algorithm.