摘要
本文提出一种基于变学习率三角基函数神经网络的线性相位4型FIR滤波器设计方法。该方法根据三角基函数神经网络与线性相位4型FIR滤波器幅频特性之间的关系,构建了一种变学习率三角基函数神经网络模型,在神经网络训练过程中引入变学习率算法自调整学习率取值,解决学习率通常依靠经验或试凑法确定带来的不确定性,提高神经网络的学习效率和收敛速度。通过训练神经网络的权值,使设计的FIR滤波器幅频响应与理想幅频响应在整个通带和阻带内的误差平方和最小。文中利用该方法对FIR高通滤波器和带通滤波器进行了优化设计,仿真结果表明了该方法设计FIR滤波器的有效性和优越性。
A novel method of designing the linear phase type-four FIR filter based on the triangle basis neural network with a variable learning rate is presented. According to the relation of the amplitude-frequency characteristics of the linear phase type-four FIR filter and the triangle basis neural network, a triangle basis neural network model with a variable learning rate is built. In the training process of the triangle basis neural network, the value of learning rate is automatically adjusted using the variable learning rate algorithm. This strategy solves the uncertainty that the learning rate usually is en- sured according to the experiences or trial and error methods. The proposed algorithm enhances the learning efficiency and the convergence rate of the neural network. By training the neural network weight, the model makes the squared sum of amplitude-frequency response error between the designed FIR filter and the ideal filter the least in the whole pass band and the cut band. The high-pass filter and band-pass filter are designed using the model in this paper. The simulation results show its availability and good performance in the design of the FIR filter.
出处
《计算机工程与科学》
CSCD
北大核心
2010年第8期141-144,共4页
Computer Engineering & Science
基金
国家自然科学基金资助项目(50677014
60876022)
高校博士点基金资助项目(20060532016)
教育部新世纪优秀人才支持计划资助项目(NCET-04-0767)
湖南省自然科学基金资助项目(06JJ2024)
关键词
三角基函数
神经网络
变学习率
高通滤波器
带通滤波器
triangle basis function
neural network
variable learning rate
high-pass filter
band-pass filter