基于RBF网络的有源噪声控制

A Study of Active Noise Control Based on RBF Neural Net works

提出一种神经自适应噪声有源控制 (ANC)的方法。应用RBF(RadialBasisFunction)网络对噪声进行有源控制。针对RBF的网络特点 ,使用递阶遗传算法确定网络参数 (连接权、隐节点中心和宽度 ) ,同时解决了网络拓扑结构的优化训练。

In this paper, a method of neuro-adaptive active nois e control (ANC) system is presented. The RBF(Radial Basis Function)is consider ed both in the modeling and control context. A hierarchical genetic algorithm fo r RBF neural networks is used to determine network parameters such as centers, w idths and connection weights. The configuration of RBF network is also establis hed at the same time during training. A feed-forward ANC system is used to can cel broadband air-condition noise in a three-dimensional propagation medium. T he developed neuro-adaptive ANC algorithm is implemented within a free-field e nvironment, and simulation results to verify its performance are presented and d iscussed.