自适应变步长正交约束盲源分离算法

将盲源分离算法通常应用到的白化预处理方法转化为权值正交约束条件下的分离算法,使得分离算法由无约束算法转变为有约束算法,消除了在估计白化矩阵时引入的误差和分离输出存在的尺度不确定性。因为算法的收敛速度和稳态误差是一对矛盾,所以结合变步长思想,提出了一种新的自适应变步长的权值正交约束盲源分离算法。该算法步长是基于分离状态的,其学习速率由信号的分离程度自适应地选取,因而能很好地解决收敛速度和稳态误差之间的矛盾。仿真实验表明,该算法比非约束算法,固定步长的权值正交约束的盲源分离算法具有更好的分离性能。

Adaptive Variable Step Size Orthogonal Constraint Blind Source Separation Algorithm

By transforming the pre -whitening into a weighted orthogonal constraint condition, the weakness that error in the estimation of pre - whitening matrix and the indeterminacy of scale is eliminated. Due to the convergence speed and the steady state error is a couple of contradiction, so a new adaptive variable step size weighted orthogonal constraint blind source separation algorithm is proposed. Because the new algorithm is based on separating degree, its learning ratio is chosen adaptively according to separating degree, therefore it can improve convergence speed and reduce the steady state error simultaneously. Computer simulations show the performance of new algorithm is superior to the unconstrained optimization model algorithms and fixed - step weighted orthogonal constraint blind source separation algorithm.