改进的粒子群算法及其非线性盲源分离

Nonlinear blind source separation using improved particle swarm optimization

采用粒子群算法与自然梯度法相结合进行非线性混叠信号盲分离。用高阶奇数多项式拟合非线性混合函数,建立非线性信号盲分离模型。同时根据粒子群算法的特点,作了改进,引入了“迁徙操作”和局部深度搜索方法。对多项式的参数用粒子群算法来求估计,然后用自然梯度法对线性去混合矩阵进行迭代。仿真结果表明,粒子群算法可以快速、有效地求得非线性混叠信号盲分离的优化解。

The nonlinear blind source separation algorithm is proposed using improved particle swarm optimization（PSO） combined with natural gradient algorithm. The model of nonlinear blind source separation（NBSS） is built which the nonlinear transfer function is simulate by the P-th order polynomial function. At the same time, based on the chatter of particle swarm optimization （PSO） , “migrating operator” and local area deep-searching is introduce into PSO. Then, the parameters of the P-th order polynomial function is estimated by PSO. Experimental results indicate that the established algorithm of PSO can quickly and effectively get optimal resolution to the nonlinear blind source separation.