基于神经元网络的自适应噪声抵消系统研究

Research on Adaptive Noise Cancellation System with Neural Network

在信号处理的研究中,自适应噪声抵消技术广泛地应用于通信、控制等领域,LMS是最常用的自适应算法,若信号通道结构比较复杂或存在非线性时,自适应滤波器的长度会增加,造成稳态失调、收敛速度降低,影响系统的性能。由于神经网络经过训练后可以很好地逼近非线性函数,因此对于平稳信号输入,采用BP神经网络构成自适应滤波器可以提高系统的抵消性能。根据神经元网络的自适应噪声抵消系统原理,通过仿真实验研究了在不同输入信噪比、不同通道函数、不同输入信号条件下系统的噪声抵消性能。实验表明BP方法噪声抵消效果显著,信噪比增益高。

Adaptive noise cancellation technique is widely used in communication,control and other areas.LMS is the most commonly used adaptive algorithm.When the noise channel structure is complex or nonlinear,the length of adaptive filter will increase,which results in increasing steady - state imbalance and low convergence rate,and affects the system performance.As the neural network can approximate to nonlinear function very well after training,an adaptive filter based on BP neural network for stationary signal can improve noise canceling performance.The basic principle of adaptive noise cancellation system based on neural network is analyzed,and noise cancellation performance of system is researched under different SNR,different channel functions and different input signals through simulations. Experiments proved that the noises can be cancelled significantly.