引入支持向量机的小波分数间隔盲均衡算法

A blind equalization algorithm based on fractionally spaced orthogonal wavelet transform with support vector machine

针对正交小波变换分数间隔盲均衡算法(FSE-WT-CMA)收敛速度慢且存在局部收敛问题,提出了一种引入支持向量机的正交小波分数间隔盲均衡算法(SVM-FSE-WT-CMA)。该算法通过对分数间隔盲均衡器的输入信号进行正交小波变换,以降低信号的自相关性,并且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,采用一小段初始数据,对盲均衡器的权向量进行初始化,然后切换至正交小波分数间隔盲均衡算法。水声信道仿真结果表明:该算法明显优于分数间隔盲均衡算法、正交小波分数间隔盲均衡算法。因此,该算法有一定的参考价值。

Due to the disadvantages in low convergence rate and local minima being found in the blind equalization algorithm based on fractionally spaced orthogonal wavelet transform and constant modulus algorithm（FSE-WT-CMA）,a blind equalization algorithm based on fractionally spaced orthogonal wavelet transform with support vector machine（SVM-FSE-WT-CMA） is proposed.This algorithm can reduce the signal autocorrelation by the orthogonal wavelet transform to the input signal of fractionally spaced blind equalizer,utilizes a short initial data segment to initialize weight vector,and turns blind equalization problem into support vector regression problem by means of the support vector machines.Then,the proposed algorithm is switched to blind equalization algorithm based on fractionally spaced orthogonal wavelet transform.Computer simulations of underwater acoustic channels indicate that the proposed algorithm obviously outperforms FSE-CMA（a fractionally spaced blind equalization algorithm based on constant modulus algorithm） and FSE-WT-CMA.