基于分组简化粒子群算法的盲源分离

传统盲源分离算法普遍存在收敛精度低和易陷入局部最优的缺点，针对上述问题，提出将蛙跳算法的分组思想应用到盲源分离算法中.该分组思想是将整个粒子群分为多组子群体，每组粒子在进行组内寻优的同时进行全局寻优，从而增加了粒子之间的差异性，可以有效避免早熟收敛.该算法以负熵为目标函数，通过对分离矩阵进行调整，使各个信号分量之间相互独立，从而完成对瞬时混合信号的盲源分离.实验仿真结果表明，提出的算法与基本的粒子群盲源分离算法相比，能有效避免早熟收敛并进一步提高收敛精度和算法的稳定性.

Blind Source Separation Based on Grouping Simplified Particle Swarm Optimization Algorithm

Traditional algorithm of blind source separation (BSS)is easy to fall into partial optimum value， and the convergence precision is low. In view of these disadvantages， the BSS method based on improved simplified particle swarm optimization algorithm was proposed， by which the whole particle swarm could be divided into several groups. Each particle was optimized while optimizing the whole area， and the difference between particles was increased. What’s more， premature convergence was avoided effectively. The negative entropy was taken as the objective function in the proposed algorithm， and the separation matrix was adjusted to separate each signal component from each other， so as to accomplish the blind source separation of instantaneous mixed signals. The simulation results show that the proposed algorithm is effective in avoiding premature convergence， and further improving convergence accuracy and algorithm stability compared with the basic particle swarm algorithm.