支持向量机初始化的常数模盲均衡算法仿真

为了提高常数模盲均衡算法的收敛速度并避免算法收敛至局部极小,提出了一种支持向量机初始化的常数模盲均衡算法.新算法采用一小段初始数据,利用支持向量机,将盲均衡问题转化为全局最优的支持向量回归问题,对盲均衡器的初始权向量进行设定,而后切换至计算量较小的常数模算法.采用浅海水声信道对新算法进行了计算机仿真,结果表明:支持向量机初始化阶段收敛速度快;切换至常数模算法后性能稳定.该算法适合应用于快衰落水声信道中通信数据的实时恢复.

Simulation of Constant Modulus Blind Equalization Algorithm Initialized by Support Vector Machines

For increasing the convergence rate of constant modulus blind equalization algorithm and avoiding local minima, a constant modulus algorithm initialed by support vector machines is presented in this paper. The new algorithm utilizes a short initial data segment, transforms blind equalization problem into support vector regression problem and sets the initial weights of the blind equalizer by means of support vector machines. Then, the, algorithm is switched to constant modulus algorithm, which has a lower computational burden. The results of computer simulation based on a shallow water channel demonstrate that the initial phase utilizing support vector machines converges fast, while the algorithm is stable after switching to the constant modulus algorithm. The proposed algorithm is suitable for real time data recovery in fast fading underwater acoustic channel.