基于状态转移法的正弦基神经网络滤波器设计

摘要

为了改善常用滤波器设计方法存在的不足,用一组正弦基函数神经网络线性组合逼近理想滤波器的振幅响应,用状态转移算法优化基函数神经网络权值,使得实际滤波器的幅度响应逼近理想滤波器的幅度响应,建立了基于状态转移算法的正弦基函数神经网络滤波器设计模型.与窗函数法设计的滤波器对比,上述模型具有更好的性能,实验结果表明,状态转移算法优化权值的方法,克服了传统基函数神经网络模型权值不易确定,收敛速度慢等缺点.%The sine basis function neural networks model optimized with state transition algorithm was proposed to improve the weakness of filter design methods which used a set of linear combination of sine basis functions to gain on the amplitude response of ideal filter.A state transition algorithm was used to optimize neural network weights.Compared with the window-function,this model has better performance.And the experimental results show that the state transition algorithm optimization weight method can overcome the shortcomings of traditional basis function such as difficult weight determination and the slow convergence speed,et.al.