基于神经网络的数字信号调制方式识别  

Modulation Types Recognition of Digital Signals Based on Neural Network

作　　者：黄付庆[1] 敖仙丹[2] 程剑俐[1] 任国春[1] 

机构地区：[1]解放军理工大学通信工程学院,南京210017 [2]重庆通信学院,400035

出　　处：《微计算机信息》2009年第4期242-244,共3页Control & Automation

摘　　要：提出了一种对经过符号成形的数字信号调制方式进行识别的新算法,利用特征参数提取与神经网络分类器相结合对8种常用数字信号进行识别。首先介绍了符号成形对信号瞬时参数提取的影响,然后选取特征量并通过所设计的神经网络分类器进行识别分类,最后计算机仿真表明利用神经网络分类器在低信噪比下能较好地区分开8种经符号成形的数字信号,且选用的L-M训练算法收敛快,适合在线分析。A new recognition algorithm to the signals symbol shaped is presented. Eight modulation types were classified by the method based on characteristic parameters extraction and neural network classifier. Firstly, the infection to parameters extraction aroused by symbol shape is introduced. Then the modulation types are classified from the feature parameters extracted and the neural network classifier designed. Lastly, the computer simulation results show that, the signals shaped could be well classified in the low SNR environment, further more the Leven-Marquart algorithm used has very speedy eonstringeney, and it could be analysised on line.

关 键 词：调制识别 符号成形 特征参数提取 神经网络分类器 

分 类 号：TN911.3[电子电信—通信与信息系统]