基于小波包变换和判决反馈RBF网络的组合均衡器

本文提出了一种基于小波包变换和判决反馈RBF网络的组合非线性均衡器的结构和算法.首先将信号进行小波包分解,再将分解后的信号分量送入带有判决反馈结构的RBF神经网络进行均衡.一方面,小波包具有很强的去相关能力,可以提高均衡器的收敛速度;另一方面,RBF神经网络具有较强的非线性模式分类能力,可降低均衡器的均方误差.在仿真实验中,针对无线通信数字信号传输过程中由于多径效应和信道衰落而产生的码间干扰(ISI)问题,比较了最小均方(LMS)算法和组合均衡器算法的均衡效果,结果表明,组合均衡算法具有更快的收敛速度,更低的误码率.

Equalizer based on Wavelet Packet Transform and Decision Feedback Radial Basis Function Neural Network

Nonlinear equalizer based on wavelet packet transform and decision feedback Radial Basis Function(RBF) neural net- work is proposed in this paper. At first signal is decomposed by wavelet packet,then decomposed signal is equalized by decision feedback RBF network. On one hand,wavelet packet has strong ability to de-correlate signal that can expedite convergence speed of algorithm;on the other hand,RBF network does well in nonlinear pattern classification that can decrease Mean Square Error(MSE) of equalizer. In simulation experiment,in order to solve Inter Symbol Interference (ISI) problem caused by muhipath effect and channel decay,equaliza- tion effect of traditional Least Mean Square (LMS) algorithm and combined equalizer are compared with each other,the later has faster convergence speed and lower bit error rate.