带有梯度加速粒子群算法的盲源分离.pdfVIP

第 27卷 第 10 计 算 机 仿 真 2010年 10月 文章编号: 1006- 9348( 2010) 10- 0112- 04 罗涛华, 张 聪 (武汉工业学院计算机与信息工程系, 湖北武汉 430023) : 研究盲源信号分离算法, 针对基于标准粒子群独立分量分析算法的盲源分离存在收敛速度慢、易于陷入局部最优值的 问题, 提出了一种带有梯度加速粒子群的盲源分离算法。以分离信号的峰度值为目标函数, 通过引入梯度信息来影响粒子 速度的更新。减小陷入局优的可能性, 当群体最优信息陷入停滞时, 对群体进行部分初始化来保持群体的活性。仿真结 果说明, 梯度信息的加入使粒子的移动更有针对性, 移动更有效率, 不仅进一步提高 PSO算法的收敛速度, 而且使改进的粒 子群算法在全局收敛性、稳定性和准确性等方面更有效。 : 盲源分离; 粒子群算法; 梯度加速; 峰度 : TN 9117 : B Blind Source SeparationUsing theParticle Swarm Optim izationAlgorithm with GradientAcceleration LUO Tao- hua, ZH NG Cong (Department ofComputer and Information Engineering, Wuhan Polytechnic University, W uhan Hubei430023, China) ABSTRACT: Traditionalparticle swarm optmi ization algorithm carried out by independent component analysismeth od is easy to fall into partial optmi um value, and the convergence precision is low. To solve this problem, a blind source separation algorithm based on particle swarm optmi ization algorithm w ith gradient acceleration is proposed, which takes the kurtosis ofm ixtures as a contrast function, and by adding gradient information to influence the update of velocities of the particles. W hen the optmi um information of the swarm is stagnant, some particles in the population are initialized again to reduce the possibility of trapping in localoptmi um. The expermi ent results reveal that applying the PSO w ith grads acceleration to BSS canmake themovement of particlemore pertinent and the movement ismore efficient. It not only mi proves the convergence speed ofPSO algorithm, but also makes the mi proved PSO algorithm is more efficient it the aspect ofglobal con