互累积量迫零法信号源盲分离

利用高阶累积量进行信号源盲分离的已有算法都需要进行复杂的矩阵代数运算，且这类算法不具备所希望的等变特性，对于病态混合矩阵的盲分离问题可能无法求解，通过利用迭代算法迫使经过非线性函数变换的混合信号互累积量矩阵对角化的方法，提出了一种新的基于高阶累积量的具有等变特性的信号源盲分离算法，该算法所采用的累积量矩阵对角化方法不依赖于混合矩阵，也不需要对累积量矩阵进行代数变换，并且所使用的迭代算法不需要对任何变量求导，因此非常简单，易于实现；同时算法还具有对未经去除均值的混合信号直接进行分离的能力。

Cross-Cumulant Zero-Forcing Blind Source Separation

Blind source separation algorithms based on higher order cumulants are all involved in complicated matrix algebra. In addition, this kind of algorithms lacks the equivariant property that is desirable in blind source separation. These algorithms may have difficulty in separating the mixtures when it is ill conditioned. A new equivariant algorithm for blind source separation was proposed by diagonalizing the cross cumulant matrix of the mixed signals that are first transformed by some nonlinear functions. The diagona lization is performed by some iterative algorithm but not matrix algebra and is not dependent on the mixing matrix. The iterative algorithm used for matrix diagonalization does not require the computation of derivatives of any variables so it is simple and easy to implement. The proposed algorithm can also perform the separation when the mean is not removed from the mixture.