基于膜计算与粒子群算法的盲源分离方法

Blind source separation method based on membrane computing and PSO algorithm

为了解决盲源分离方法收敛速度慢、分离性能不高的问题,提出一种基于膜计算(Membrane Computing,MC)和粒子群算法(Particle Swarm Optimization,PSO)的盲源分离方法。算法以分离信号负熵作为粒子群的适应值函数,将粒子均匀分布到各基本膜中,将各基本膜内最优位置输出到表层膜并选择适应值最小的最优位置作为群体最优位置,通过粒子自身最优位置和群体最优位置对种群粒子进行速度和位置的更新。粒子群最优解调整盲源分离的步长函数,进行信号的分离。提出的算法简化了惯性权重取值问题,保证了PSO算法局部搜索的精度,满足了全局搜索的多样性。仿真实验和实例应用表明,提出的算法可以很好地分离混合信号,并且能避免PSO算法的早熟收敛问题,具有更快的收敛速度和更优异的分离性能。

In order to solve problems of slower convergence and lower separating performance of the existing blind source separation methods,a new method based on membrane computing(MC)and particle swarm optimization(PSO)was proposed.The separated signals’neg-entropy was taken as the fitness function of PSO.Particles were uniformly distributed into each elementary membrane.The velocity and position of population particles were updated with a particle own optimal position and the population global optimal position.The optimal solution to PSO was used to adjust step function of blind source separation and then separate signals.The proposed algorithm simplified inertia weight’s choosing to ensure the accuracy of PSO local search and satisfy the variety of global search.The simulation and actual application showed that mixing signals can be well separated with this new method and the premature convergence problem of PSO can be avoided;this new method has a faster convergence speed and a more excellent separating performance.