基于混合遗传优化的正交小波变换盲均衡算法

摘要

Constant modulus algorithm (CMA) has slow convergence speed and easily im- merges in partial minimum owing to the lack of initialization theory. Aiming at these disadvantages, hybrid blind equalization algorithm based on genetic algorithm and wavelet transform (GAWT-CMA) is proposed to improve the performance of CMA. The proposed hybrid algorithm is formed by embeding WT-CMA in normal genetic algorithm. It uses the global convergence of genetic operator to conduct a macro-search, and conducts micro-search with WT-CMA. It obtains a better initialization weight vector in overall range with a little bit of data by the hybrid algorithm, and then gets globally optimal solution by WT-CMA. Computer simulation with underwater acoustic channel shows that GAWT-CMA' s convergence is about 9 000 times faster than CMA, about 3 000 times faster than WT-CMA, and its steady-state error is smaller than CMA by 3 dB, smaller than WT-CMA by about 1 dB.%常数模算法(CMA)收敛速度慢,初始权向量的确定缺乏理论依据,容易陷入局部极小值.针对这些问题,在正交小波变换盲均衡算法(WT-CMA)的基础上,提出了基于混合遗传优化的正交小波变换盲均衡算法(GAWT-CMA).该算法在常规遗传算法的父代和子代之间嵌入WT-CMA形成混合算法,利用遗传算子的全局收敛性进行宏观搜索,用WT-CMA进行局部搜索.由混合算法用少量数据进行权向量的优化,在全局范围内获得较好的初始权向量,再用WT-CMA收敛到全局最优值.水声信道的仿真结果显示,GAWT-CMA完成收敛比CMA快约9 000次,比WT-CMA快约3 000次,比GA-CMA快约1 000次,稳态误差对比CMA小约3 dB,比WT-CMA小约1 dB,比GA-CMA约小0.5 dB.