一种新的双模式神经网络盲均衡器设计与仿真

Design and Simulation on Novel Dual-Mode Neural Network Based Blind Equalizer

在通信系统设计中,常采用盲均衡器来抑制带限信道导致的码间干扰。但传统的常数模算法(CMA)以及基于CMA的双模式算法对于多进制正交调幅信号(QAM)存在较大的误判,收敛后QAM系统性能较差等不足之处。在修正常模算法(MCMA)的基础上,针对QAM信号为多模信号的特点,采用多模算法(MMA)与修正判决引导算法(MDD)结合的双模式算法,并引入可准确模拟信道逆系统的多层感知机结构,得到了一种新的基于双模式算法的多层感知机结构神经网络盲均衡器,利用新算法调整神经网络参数,并且利用重置模块跟踪信道变化。仿真结果表明,新算法调整的神经网络盲均衡器双模式盲均衡器在稳态MSE、收敛性方面都有所提高,并具有抵抗信道突变的能力,为通信系统设计提供参考。

In modem communication system, blind equalizer is widely used to restrain the inter - symbol interfer- ence （ISI） caused by non- ideal character of channel. However, the traditional constant modulus algoIithm（CMA） and those dual -mode algorithms based on CMA often cause false estimation when it means to m -ray QAM signal. Thus, for the multi -modulus feature of m- ray QAM signals, based on modified constant modulus algorithm（ MC- MA） , a new dual - mode with multi - modulus algorithm（MMA） and modified decision - directed algorithm（MDD） was adopted in the multilayer perception structure. The new equalizer adjusted the parameters with the new algorithm and used the reset device to track the burst channel. Results show that the new dual - mode blind equalizer based on neural network has fast convergence speed, lower steady - state MSE and the capability to track the burst channel.