中文摘要：

现有的盲信源分离方法大多以信源统计独立为前提条件.针对相关信源的瞬时混合,提出了一种采用＂两步法＂对其进行盲分离的方法.首先,对两传感器观测信号比值的方差进行比较,方差值相等的相继采样时刻就是只有一个源信号单独存在的时刻.这些时刻的观测信号矢量就是对混合矩阵中与该源信号对应的列矢量的估计,利用这一性质可以估计出混合矩阵.然后,由估计出的混合矩阵得到分离矩阵,从而恢复出源信号.计算机仿真试验证明了该方法的正确性.

英文摘要：

The existing methods of blind source separation are mostly based on the precondition of independence of sources. A new blind source separation method is proposed in this paper, which is a two- stage approach and applicable to instantaneous mixtures of dependent sources. Firstly, variances of the ratio between two sensors are compared. Sequent sampling instants, whose varianees are equal, are the instants at which only one source exists. The vectors of observation at these instants, are the estimate of the corresponding columns at the mixing matrix. By using this property, the mixing matrix can be estimated correctly. Then, the de-mixing matrix is obtained and sources are recovered. Simulation results illustrate the correctness of the algorithm.