盲信号分离的自稳定算法

Self-stabilized Algorithm for Blind Signal Separation

在传统盲分离算法的迭代过程中,为了保证分离矩阵的行正交性,都要对其进行反复的正交化,以避免算法的不收敛。针对这个问题,笔者主要研究了在StiefelManifold上的盲分离算法,该算法使分离矩阵不需要每一步都进行正交化。仿真表明,本文提出的新算法具有很好的稳定性。

For conventional blind separation algorithms, the rows of the demixing matrix need to be periodically re-orthonormalized during the iteration in order to avoid divergence. To this problem, we study the blind source separation algorithm in the Stiefel manifold in this paper, so that the rows of the demixing matrix do not need to be periodically re-orthonormalized. Through simulation,it is shown that the new algorithm has superior stability.