基于深度神经网络的数字信号调制类型自动识别方法  

Automatic Modulation Recognition of Digital Signals Based on Depth Neural Network

作　　者：丁锡龙[1] 金乾坤[2] 

机构地区：[1]潍坊科技学院中印计算机软件学院,山东寿光262700 [2]北京理工大学软件学院,北京100081

出　　处：《湘潭大学自然科学学报》2017年第3期70-74,共5页Natural Science Journal of Xiangtan University

基　　金：山东省教育厅项目(J16LN51);潍坊市科技局项目(201301161)

摘　　要：针对在莱斯衰落信道下传统神经网络数字信号调制识别准确率会有所降低的问题,提出了一种基于深度神经网络技术的数字信号调制类型自动识别方法.该方法从接收信号样本中提取了多组特征参数作为分类器输入,用于识别4ASK,BPSK,QPSK,8PSK,16QAM和32QAM这些数字调制信号.此外,分类器采用了具有三个隐藏层的全连接深度神经网络结构.仿真结果表明,分别在加性高斯白噪声和莱斯衰落信道下,提出的方法与现有分类方法相比具有较好的性能改进,尤其是在具有较高多普勒频率时.According to the problem that the accuracy of digital signal modulat ion and recogni tion of tradi -tional neural network will be reduced in Rician Fading Channel, a method for automat ic ident i f icat ion of dig-i tal signal modulat ion depth based on neural network technology is proposed in this method, a set of feature parameters are extracted f rom the received signal samples as the input of the classif ier to identi fy the 4ASK, BPSK, QPSK, 8PSK, 16QAM and 32QAM. In addi t ion, the classif ier uses a fully connected neural net-work archi tecture wi th three hidden layers. The simulat ion results show that the proposed method has bet-ter performance compared wi th the exist ing classif icat ion methods in the addi t ive Gauss whi te noise and Ri -cian Fading Channels, especial ly when the high frequency of the is high.

关 键 词：调制方式识别 深度神经网络 高斯白噪声 衰落信道 多普勒频移 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]