神经网络在通信系统回音对消中的应用  

Application of neural network in echo cancellation of communication system

作　　者：余尤好[1] 

机构地区：[1]莆田学院电子信息工程系,福建莆田351100

出　　处：《武汉工程大学学报》2012年第9期70-74,共5页Journal of Wuhan Institute of Technology

基　　金：福建省教育科学"十二五"规划2012年度常规课题(FJCGGJ12-034)

摘　　要：通信系统中的回音影响通话质量,造成通话困难.为控制通信系统回音的影响,利用自适应线性神经网络建立回音对消系统,求出话音经过传输过程发生非线性变化后的结果.对神经网络学习速率和训练次数进行调节,获得回音信号的最佳估计值,原始信号与之相减实现回音对消.实验时分别对通信系统直接回音和间接回音进行消除,并利用MATLAB软件绘出对应信号的频谱图和语谱图,以便观察分析.仿真结果表明,消除回音后有用的语音信号得到增强,回音被有效削弱,通话具有较高的信噪比,人耳感受基本没有回音.Echo of the system affects the speech quality seriously and makes it difficult to communicate. Adaptive linear neural network was used to control echo effects of communication system. The echo cancellation system was established by adaptive linear neural network, and nonlinear variation results of voice was obtained. Adjusting the learning rate and the number of training, the best estimate of the echo signal was obtained by adaptive linear neural network, subtracting it from the original signal, and then echo cancellation was realized. Direct echo and indirect echo were eliminated from communication system in the experiments. In order to observe and analyze the results, MATLAB software was used to draw the corresponding signal spectrum and spectrogram. The simulation results show that the signal is enhanced after echo cancellation and echo is effectively reduced; the call has a high signal-to-noise ratio and people feel no echo.

关 键 词：回音对消 滤波器 神经网络 语音通话 

分 类 号：TN912.3[电子电信—通信与信息系统]