论文部分内容阅读
针对多个矩阵近似联合对角化盲分离问题,提出一种新的非正交近似联合对角化算法.首先采用罚函数法将联合对角化的非线性约束优化模型转化为无约束优化模型;其次将粒子群优化算法引入无约束优化模型中实现目标函数的最优化,从而完成矩阵组的联合对角化.分析了惩罚因子的更新策略及算法的收敛性能,并设计仿真实验进行对比分析以检验算法解决实际盲分离问题的能力.
In order to solve the problem of joint approximation with multiple matrices, a new non-orthogonal approximation combined diagonalization algorithm is proposed. First, the penalty function method is used to convert the combined diagonalized nonlinear constrained optimization model into an unconstrained optimization model Secondly, the particle swarm optimization algorithm is introduced into the unconstrained optimization model to optimize the objective function, so as to complete the joint diagonalization of the matrix group.Analyzing the updating strategy of the penalty factor and the convergence performance of the algorithm, the simulation experiment is designed for comparative analysis To test the ability of algorithms to solve the problem of actual blind separation.