基于稀疏贝叶斯学习的DOA估计  

DOA Estimation Based on Sparse Bayesian Learning

作　　者：董天宝[1] 汪海兵[1] 曾芳玲[1] 

机构地区：[1]电子工程学院,合肥230037

出　　处：《火力与指挥控制》2015年第3期42-45,共4页Fire Control & Command Control

基　　金：国家自然科学基金(61272333);安徽省自然科学基金资助项目(1208085 MF94)

摘　　要：针对基于l1范数约束的稀疏表示DOA(Direction Of Arrival)估计算法对初始参数较为敏感的问题,提出了一种基于稀疏贝叶斯学习的DOA估计算法。首先通过信号来波方向的空间采样构造冗余字典,将阵列信号处理中的DOA估计信号模型转化为压缩感知中的稀疏重构信号模型。然后基于经验贝叶斯推理的方法,将待估计的稀疏系数值用方差未知的联合高斯分布描述,而未知的方差值决定了待估计系数的稀疏性。通过观测数据估计得到未知的方差,进而得到信号的DOA估计值。仿真结果表明,提出的算法有较高估计精度,并且对非相干信源和相干信源都具有较好的估计性能。A Direction-Of-Arrival （DOA） estimation algorithm based on sparse Bayesian learning is proposed,which is less sensitive to initial parameters than other DOA estimation algorithms based on l1-norm. First,a DOA estimation model is transformed to one sparse signal reconstruction model by constructing a redundant dictionary from spatial samples of the interested signals. Then, sparse parameters are represented by a union Gaussian distribution with unknown variances using empirical Bayesian analysis method,and the unknown variances determine the sparsity of the parameters. Finally,the unknown variances are estimated from observation data and the DOAs of the interested signals are estimated. Simulation results show that the proposed algorithm can estimate the DOAs of signals with high accuracy, and it is effective for both coherent signals and noncoherent signals.

关 键 词：波达方向 稀疏表示 贝叶斯学习 

分 类 号：TN911.6[电子电信—通信与信息系统]