机构地区: 江南大学通信与控制工程学院
出 处: 《现代电子技术》 2009年第11期167-170,共4页
摘 要: 提出一种基于模拟退火神经网络设计FIR数字滤波器的方法,是对用神经网络设计方法的一种改进。由于线性相位FIR数字滤波器的幅频特性是有限项的傅里叶级数,因此构造了一个三层余弦基神经网络模型,并用模拟退火算法进行了优化,然后给出了高阶滤波器优化设计的实例。仿真表明经优化设计后的滤波器具有更好的性能和更稳定的效果。 A method of FIR digital filter design based on SA neural network is proposed, which improves the method of using neural network. Amplitude - frequency FIR filter with linear phase is characteristic of finite Fourier progression, so a neural networks of Cosine basis with three layers is set up,and then optimizes it with SA,and also gives the examples of the optimized design on the high- order FIR digital filters. Experiment results show that through SA, the filter has better effect and stability.
关 键 词: 余弦基神经网络 数字滤波器 模拟退火 幅频特性 优化设计
领 域: [自动化与计算机技术] [自动化与计算机技术]