一种基于吉布斯抽样的MUSIC频率-方位联合估计算法  

New Gibbs Sampling Freq-DOA Joint Estimator Based on MUSIC Method

作　　者：金勇[1] 黄建国[1] 谢达[1] 

机构地区：[1]西北工业大学航海学院,陕西西安710072

出　　处：《系统仿真学报》2008年第12期3100-3103,共4页Journal of System Simulation

基　　金：国家自然科学基金(60572098);航空科学基金(05F53027)

摘　　要：将马尔可夫蒙特卡罗方法与MUSIC方法估计相结合,提出一种基于吉布斯抽样的频率-方位联合估计新方法(MUSIC FREQ-DOA joint Estimator Based on Gibbs Sampling,简称Gibbs-MUSIC)。并用该方法联合估计多个目标的频率和方位。这种方法将MUSIC方法的谱函数作为频率和方位的联合概率密度函数,并采用马尔可夫蒙特卡罗(MCMC)吉布斯抽样方法对该联合概率密度函数进行采样。仿真实验显示在目标个数较少时,该方法不仅保持了常规MUSIC方法的高分辨能力,而且降低了计算量,减少了存储量。A new Gibbs sampling DOA estimator based on MUSIC method （Gibbs-MUSIC） is proposed to joint estimate the frequency and directions of multiple sources. The method regards the power of MUSIC spectrum function as target distribution up to a constant of proportionality, and uses Gibbs sample, one of the most popular Markov Chain Monte Carlo （MCMC） technique, to sample from it. Simulations show that the new method not only possesses the performance of high-resolution frequency and direction finding in conventional MUSIC method but also provides a less computation and storage costs than that of conventional MUSIC method under the condition where the number of signal sources is small.

关 键 词：MUSIC方法 吉布斯抽样 马尔可夫链蒙特卡罗(MCMC)方法 联合估计 计算复杂度 

分 类 号：TN911.7[电子电信—通信与信息系统]