基于相位变换和GhostNet-门控循环单元的自动调制识别方法  

Automatic Modulation Recognition Method Based on Phase Transformation and GhostNet-GRU

作　　者：陈昊 郭文普 康凯 CHEN Hao;GUO Wenpu;KANG Kai(Rocket Force University of Engineering,Xi’an 710025,Shaanxi)

机构地区：[1]火箭军工程大学,陕西西安710025

出　　处：《火箭军工程大学学报》2024年第4期86-92,共7页Journal of Rocket Force University of Engineering

摘　　要：针对信号调制方式低信噪比条件下识别准确率不高的问题,提出了一种由相位变换、GhostNet、压缩与激励网络(Squeeze and Excitation Network,SENet)、门控循环单元(Gated Recurrent Unit,GRU)和深度神经网络组成的模型,用于自动调制识别接收信号。首先,采用基准数据集RML2016.10a和RML2016.10b同相正交数据作为模型输入;其次,构建识别模型,其中,相位变换用于降低相位偏移对调制识别的影响,GhostNet和GRU分别用于提取调制信号的空间特征和时间特征,SENet用于对特征图权重进行调整;而后,通过深度神经网络进行分类;最后,对所提模型进行了训练及测试。实验结果表明:与现有模型CGDNet、CLDNN、IC-AMCNet、MCLDNN和LSTM相比,所提出模型显著降低了参数量,有效提升了低信噪比条件下的识别准确率,平均识别准确率分别达到62.30%和64.45%。To address the problem of low recognition accuracy under low signal-to-noise ratio(SNR),a model consisting of phase transformation,GhostNet,Squeeze and Excitation network(SENet),gated recurrent unit(GRU),and deep neural network was proposed,by which received modulated signals could be automatically recognized.Firstly,the benchmark datasets RML2016.10a and RML2016.10b,as well as in-phase quadrature(I/Q)data were used as input for the model.Secondly,a recognition model was constructed,where phase transformation was used to reduce the impact of phase shift on modulation recognition,GhostNet and GRU were used to extract spatial and temporal features of modulation signals and SENet was used to adjust weights of feature maps.Then,deep neural networks were used to classify the processed signal.Finally,the proposed model was trained and tested.Experimental results showed that compared with existing models CGDNet,CLDNN,IC-AMCNet,MCLDNN and LSTM,the proposed model significantly reduces the number of parameters and effectively improves the recognition ac-curacy under low SNR,with average recognition accuracies of 62.30%and 64.45%,respectively.

关 键 词：自动调制识别 深度学习 相位变换 GhostNet 门控循环单元 

分 类 号：TN911.3[电子电信—通信与信息系统]