FIR带通滤波器优化设计研究

Optimal Design Study about the FIR Band-Pass Filters

提出了一种基于 BP学习算法的正弦基函数神经网络模型 ,给出了该神经网络模型的收敛性条件 ,为神经网络训练的学习率选取提供了依据。根据本文提出的优化设计算法 ,作者详细研究了 FIR线性相位带通滤波器优化设计实例。研究结果表明 ,本文设计的 FIR带通滤波器 ,基阻带衰耗特性好 ,最小衰耗分别在 10 0分贝和 14 0分贝以上 ,这是任何其它优化设计方法难以实现的。

This paper presents the model of sine basis functions nerual networks based on BP learning algorithm, and offers the convergence condition of the neural networks algorithm. The convergence condition provides the evidence for selecting learning ratio in traning neural networks. Author studies in detail the optimal design examples about the FIR band pass filters with a linear phase according to the optimal design algorithm presented in the paper. The study results show that the attenuation performance of the stop band of the FIR band pass filters designed in the paper is excellent, which is over 100dB and 140dB respectively. The neural network algorithem is not only effective, but also better than other optimal algorithms.