基于信息论准则的信源个数估计算法改进  

Improvement of source number estimation based on information theoretic criteria

作　　者：叶中付[1] 向利[1] 徐旭[1] 

机构地区：[1]中国科学技术大学信号统计处理研究室,安徽合肥230027

出　　处：《电波科学学报》2007年第4期593-598,共6页Chinese Journal of Radio Science

基　　金：安徽省人才开发基金(2004Z025)资助项目

摘　　要：基于信息论准则(AIC准则,MDL准则)的信源数估计算法及其改进算法在低快怕数和低信噪比情况下大多得不到较好的估计性能。从最基本的MDL准则出发,充分考虑AIC准则与MDL准则估计的优缺点,通过引入影响因子对准则的代价函数进行改进,这样的改进在低快怕数与低信噪比下都能得到较好的估计性能,且具有估计一致性、算法复杂度低等优点。仿真试验验证了该改进算法的有效性。The algorithms of source number estimation and their improved algorithms based on information theoretic criteria （Akaike Information Theoretic Criteria and Minimum Description Length Criteria）, couldn＇t get a good performance both in the low SNR and short snapshots. From the basic MDL criteria, and considering the advantages and shortfalls of the AIC criteria and MDL criteria, a new im- proved algorithm is presented in this paper. It modifies the penalty function of the criteria by introducing the effect factor, and thus shows a good performance both in the low SNR and short snapshots. Furthermore, it shares the advantages of strong consistency and low complexity. The simulation demonstrates the efficiency of the improved algorithm.

关 键 词：信源数估计 MDL准则 AIC准则 阵列信号处理 

分 类 号：TN911.23[电子电信—通信与信息系统]