动态目标函数下的神经网络盲均衡仿真研究  

Blind Equalization Based on FNN Using Dynamic Object Function

作　　者：肖瑛[1] 李振兴 董玉华[1] 

机构地区：[1]大连民族学院机电信息工程学院 [2]大连91550部队94分队

出　　处：《系统仿真学报》2009年第14期4331-4334,共4页Journal of System Simulation

基　　金：大连民族学院人才引进科研启动基金项目(20066105)

摘　　要：盲均衡算法中常数模(CMA)和判决引导(DD)模式适时切换可以有效提高算法收敛速度和均衡性能,但是切换时机难以选择,同时在信道突发干扰条件下无法保证算法的稳健性。为此结合CMA和DD算法代价函数构造了一种动态目标函数,目标函数随着均衡算法迭代次数进行调整,以前馈神经网络(FNN)作为盲均衡器实现信道均衡,并以移动窗剩余误差变化率为判据对目标函数进行适时重置克服信道可能的突发干扰。计算机仿真表明了算法的有效性。A dynamic object function which takes advantage of DD algorithm and CMA algorithm for blind equalization basedon FNN was proposed, and it＇s updateby iterative times. By this means it overcame the difficulty to choose the time when the algorithm switched CMA to DD for dual-model blind equalization. Instantaneous error was used as the criterion to reset the object function to cope with the burst interference of channel, the steadiness of this method was ensured as well. Simulation results show that the proposed approach has higher convergence rate and better performance than direct blind equalization based on FNN.

关 键 词：常数模算法 判决引导算法 神经网络 盲均衡 动态目标函数 

分 类 号：TN911.5[电子电信—通信与信息系统]