改进的变步长维纳系统盲源分离方法

An Improved Variable Step Size Wiener System Blind Source Separation Method

基于非线性盲源分离的维纳系统算法中,采用固定步长导致算法的收敛速度和稳态误差之间存在矛盾,直接影响分离算法的性能。为了解决该问题,提出了基于非线性函数的变步长维纳系统盲源分离方法。该方法将更新的步长以非线性函数的形式引入到分离算法中,使得稳态时参数更新的步长尽可能小,以避免发生振荡。变步长算法在分离过程中的每次更新都会使步长自动进行合理的调整,使得收敛速度提高了53%,误差减小了45%。实验仿真表明,相对原算法,提出的维纳系统盲源分离方法可以更好地分离出信源信号,而且具有较小的误差和较快的收敛速度。

In the nonlinear blind source separation algorithm of Wiener system,the fixed step size leads tothe contradiction between convergence speed and steady-state error of the algorithm,which directly affectsthe performance of the separation algorithm. To solve the problem,an algorithm based on variable step sizeblind source separation method of Wiener system is proposed in this paper. This method that updates thestep length in the form of a nonlinear function is introduced into the separation algorithm,makes the steady-state parameters to update the step size as small as possible,so as to avoid oscillating. On the contrary,step increases,speeds up the convergence,increases steady-state error. The variable step size algorithm inthe process of separating each update makes step automatically and reasonably adjust,improves the per-formance of Wiener system blind source separation algorithm. Simulation shows the blind source separationmethod of Wiener system in this paper can separate source signals better with lower errors and faster con-vergence speed in comparison with the original algorithm.