用RBF神经网络逼近Volterra级数实现非线性噪声的消除

Using RBF neural network to approach volterra's series realizes the nonlin earity noise cancellation

Volterra级数展开式的非线性滤波器由于综合地利用了线性和非线性项,因而比其他非线性滤波器具有更好的性能。但随着Volterra滤波器阶数的增加,对应的滤波系数个数呈几何级数增长,实现非常困难。本文利用RBF神经网络具有非线性函数的逼近能力和其局部逼近网络学习速度快的优点,用RBF神经网络逼近Volterra滤波器以实现非线性噪声消除。Matlab仿真结果表明了此方法的有效性。

With used linearity and the nonlinearity item,the Volterra's series filter has better capability.But with the increase of the Volterra filter order,the number of filter coefficient assumes the geometric series increasment.It's very difficult to realize the fil-ter.RBF neural network has the capability of approaching the nonlinear function and the quick study speed of local network.This paper uses RBF neural network approaching the volterra filter to cancel nonlinearity noise.Matlab simulations indicate the effective-ness of this method.