基于BP神经网络算法优化FIR数字滤波器的研究

本文将BP神经网络算法应用于FIR线性相位数字滤波器的设计中，通过训练神经网络的权值，得到滤波器的单位冲激响应。利用此方法和传统的窗函数法分别设计了50hz带阻滤波器，通过对两种滤波器幅频特性进行比较，发现神经网络滤波器吴有更好的阻带衰减效果，并且无过冲波动，性能更优。

The design of a FIR linear phase filter based on BP neural network algorithm is introduced, get the unit impulse response of the filter by training the wieght of the BP neural network. A 50Hz band-stop filter has been designed by this method and the windowruction method,which has been founded that the filter designed by BP neural network is better at attenuate the stopband by comparing with two kinds of method,and no over-impulse.This method has the better performance.