基于倒三谱初始化的Bussgang类盲均衡算法研究

水声通信信道带宽非常有限。与自适应均衡算法相比,盲均衡算法不需要训练序列而节省了带宽,特别适合于高速水声通信和多点通信。Bussgang 类盲均衡算法计算简单,但收敛速度慢,且对权向量的初始化很敏感。不恰当的初始化会造成算法收敛到局域极小值。多谱类盲均衡算法收敛速度快,且是全局最优的,但计算量大,不利于硬件实现。本文利用倒三谱,根据倒三谱累积量方程,由一般数据计算得到倒谱参数,从而直接估计出均衡器参数,对Bussgang类算法进行初始化,大大加快了算法的收敛速度,保证了算法收敛到最优。通过对浅海信道均衡的计算机仿真,证明了该方法的有效性。

On the Bussgang Class Blind Equalization Algorithms Based on Tricepstrum Initialization

The bandwidth of the underwater communication channel is very limited. Compared with the adaptive equalization algorithms, blind equalization needs no training sequence, so that save lots of band width, which makes it very suitable for the scenarios of high data rate underwater communication and multipoint communication. Blind equalization algorithms of Bussgang class have the advantage of computing efficiency and disadvantages of slow convergence rat sensitivity to initialization conditions. Improper initialization will lead the algorithms converge to local minimum. The polyspectra algorithms converge fast and always converge to global minimum, but the computational complexity made them unsuitable for hardware implementation. A new method based on tricepstrum initialization is discussed in this paper. According to the cepstrum cumulant equation, the cepstrum parameters are calculated from a data set, and the tap weights estimated directly for the initialization of Bussgage class algorithm. This method accelerates the convergence rate of the Bussgang class algorithms, and ensures them converging to global minimum. Through computer simulation of shallow water channel equalization, the efficiency of the method is proved.