一种神经网络非线性噪声消除方法

针对阶数的增加,Volterra滤波器的滤波系数个数呈几何级数增长,实现困难的问题,提出采用基于Levenberg-Marquardt(LM)算法的BP神经网络逼近Volterra滤波器,实现神经网络非线性滤波,使计算简化。采用的方法是,先离线调整Volterra滤波器的系数,再用调整好的Volterra滤波器监督训练LM-BP神经网络,然后用训练好的LM-BP神经网络进行非线性自适应滤波。仿真实验结果表明,LM-BP神经网络滤波器较其学习导师———Volterra滤波器具有更好的噪声滤除效果。

Approach of nonlinear noise cancellation based on neural network

The number of the Volterra filter coefficients assumes a geometric series increase with the increase of the filter order,and it is very difficult to realize the filter.The BP neural network based on Levenberg-Marquardt(LM) algorithm is employed to approach the Volterra filter and to cancel nonlinear noise.The algorithm of the neural network is simple.First the Volterra filter coefficients are adjusted in an off-line manner.Then the LM-BP neural network is trained under the supervision of the Volterra filter.Finally the LM-BP neural network trained is used in adaptive nonlinear noise filter.The results of the simulation experiments show that the LM-BP neural network filter has a better result in noise cancellation than the Volterra filter.