作　　者： ; ;

机构地区： 华中科技大学机械科学与工程学院

出　　处： 《数据采集与处理》 2008年第1期1-7,共7页

摘　　要： 通常盲源分离以及源数估计均要求传感器数必须大于或等于源数,对于实际复杂系统来说,这个条件很难满足。为此,提出一种基于功率谱密度的源数估计方法。该方法最大特点在于,无论传感器数与源数关系怎样,根据观测信号的功率谱密度函数的比值,得到功率谱密度矩阵,通过比较矩阵各列向量,可准确估计不相关源或独立源的源数。对于相关源,可估计源数的上下界。理论分析和仿真实验证明了该方法的有效性,同时还仿真了噪声对源数估计的影响。 Usually in blind source separation （BSS） and estimation of the source number, the condition is very difficult to satisfy that the number of sensors must be more than or equal to the number of the sources for complex systems. A new estimation method based on power spectral density（PSD） is presented. Based on the ratio of PSD of the observation signals, the PSD matrix is obtained. And the column vectors of the matrix are examined to estimate the sources number. The distinct property of the method is that no matter the number of sensors is more than that of the sources or not, the number of the uncorrelated or independent sources can be accurately determined and the bound of the number of the correlated sources can be estimated. The effectiveness of the proposed method is verified by theoretical analysis and simulation. Finally, the effect of the noise on the estimation of the source number is analyzed.

关 键 词： 盲信号 源数 估计 功率谱密度函数

领　　域： [电子电信] [电子电信]