|
| 摘要: |
| 为有效提高有限冲激响应FIR(FiniteImpulseResponse)高阶数字滤波器优化设计速度和精度,根据FIR线性相位滤波器的幅频特性,提出了一种基于激励矩阵为Hd-CTW的神经网络算法。该算法的主要思想是用神经网络算法优化设计的FIR滤波器的幅频特性与理想滤波器的幅频特性在整个通带和阻带范围内的误差平方和为最小,算法不涉及逆矩阵运算。为了保证该算法的收敛性,提出并证明了神经网络算法的收敛性定理,为神经网络学习率的选择提供了理论依据。该算法的主要特点是可实现样本集数据的并行训练,有效提高了计算速度。为了验证该算法的有效性,给出了多通带FIR高阶数字滤波器优化设计实例,仿真结果表明了该算法具有高的计算精度和快的计算速度。 |
| 关键词: 神经网络,FIR高阶数字滤波器,优化设计 |
| DOI： |
| 分类号:TN713 |
| 基金项目:湖南省自然科学基金项目(06JJ5011)~~ |
|
| Optimal design of high order multi-band-pass FIR digital filter |
|
LI Si AN Wei-ke
|
| Abstract: |
| To improve the design speed and precision of high-order FIR(Finite Impulse Response)digital filter,a neural network algorithm based on the activation matrix Hd-C T W is presented,which makes the square sum of amplitude-frequency response error between the designed FIR filter and the ideal filter least in the whole pass band and cut band.The invert matrix operation is not involved in the algorithm and its convergence theorem is presented and proved,which provides the theoretical basis for setting the learning rate of neural network.The sample data are trained in parallel to speed up the calculation.Examples of the optimal FIR digital filter design are given and the simulative results show its high precision and fast convergence rate. |
| Key words: neural network,high-order FIR digital filter,optimal design |
