基于径向基函数神经网络的自适应滤波器

自适应滤波器在信号检测、信号恢复、数字通信等领域中被广泛应用。传统的自适应滤波器主要在时域中实现。通常采用算法简单、稳健性好的自适应LMS算法。但LMS算法对输入信号的自相关矩阵具有很强的依赖性,因而自适应率不高。本文提出利用RBF神经网络实现的自适应滤波,并将其用于语音除噪和脉象信号的除噪。仿真结果表明该方法具有良好的非线性噪声抑制能力。

Adaptive Filter Based on Radial Basis Function of Neural Network

Adaptive filters are commonly used in the field of signal detection,signal restore and digital communication.The classical adaptive filters are principally implemented in the time domain.The classical algorithm of adaptive filter is the least mean square(LMS) algorithm.The LMS algorithm has the characteristic of being simple and stable.Due to the fact that LMS algorithm highly dependents on the autocorrelation matrix,the algorithm has low adaptive ratio.In this paper,we improve the method by using the radial basis neural network to realize the adaptive filter.After using it in the field of removing noise in the voice signal and controlling noise in the pulse signal,we find the method has good performance in controlling nonlinear noise.