基于残差生成对抗网络的调制识别算法

doi:10.12305/j.issn.1001-506X.2022.06.30

摘要/Abstract

摘要：

针对小样本条件下通信信号识别准确率不高、网络训练困难的问题, 本文提出一种基于残差生成对抗网络的调制识别算法。首先, 设计一种以Leakyrelu作为隐藏层激活函数的新残差单元, 使得网络对输入为负值的数据也可以进行梯度计算; 然后, 将新残差单元组成的残差网络和卷积神经网络作为本文算法的基本网络结构, 使用卷积步幅为1的非对称小卷积核, 更好地提取信号的边缘特征信息; 最后, 用Dropout代替池化操作, 并选择Adam梯度优化算法以交替迭代方式完成网络训练。仿真实验结果表明, 小样本条件下, 残差生成对抗网络算法复杂度明显降低, 信噪比(signal to noise ratio, SNR)在0 dB以上时, 对10种调制信号的识别准确率可以达到91%, 验证了所提方法的有效性。

关键词: 残差网络, 生成式对抗网络, 调制识别, 小样本

Abstract:

Aiming at the problem of low accuracy of communication signal recognition and difficult training of network in few-shot learning, a modulation recognition algorithm based on the residual generative adversarial network is proposed. Firstly, this algorithm designs a new residual block unit using Leakyrelu as the activation function of the hidden layer, which can also calculate the gradient for the negative input data. Then, the residual network composed of the new residual units and the convolutional neural network are used as the basic network structure of this proposed algorithm. Moreover, we apply the asymmetric small convolutional kernel with a convolutional step of 1 to better extract the edge feature information of the signal. Finally, Dropout is adopted to replace the pooling operation, and Adam gradient optimization algorithm is selected to complete the network training in an alternate iteration mode. Experimental results show that under the condition of a few-shot learning, the recognition accuracy of the residual generation adversarial network algorithm for 10 modulated signals can reach 91% when the signal to noise ratio (SNR) is higher than 0 dB, which verifies the effectiveness of the proposed method.

Key words: residual network (Resnet), generative adversarial network, modulation recognition, few-shot

中图分类号:

TB975

秦博伟, 蒋磊, 许华, 齐子森. 基于残差生成对抗网络的调制识别算法[J]. 系统工程与电子技术, 2022, 44(6): 2019-2026.

Bowei QIN, Lei JIANG, Hua XU, Zisen QI. Modulation recognition algorithm based on residual generation adversarial network[J]. Systems Engineering and Electronics, 2022, 44(6): 2019-2026.

图/表 15

图1

图2

图3

图4

表1

图5

图6

图7

图8

表2

图9

图10

图11

表4

表5

参考文献 22

1	WU H , LI Y X , ZHOU L , et al. Convolutional neural network and multi-feature fusion for automatic modulation classification[J]. Electronics Letters, 2019, 55 (16): 895- 897. doi: 10.1049/el.2019.1789
2	TANG B , TU Y , ZHANG S Y , et al. Digital signal modulation classification with data augmentation using generative adversarial nets in cognitive radio networks[J]. IEEE Access, 2018, 6, 15713- 15722. doi: 10.1109/ACCESS.2018.2815741
3	童年, 李立春. 非重构压缩样值的MPSK信号最大似然调制识别[J]. 信息工程大学学报, 2015, 16 (5): 579- 583. doi: 10.3969/j.issn.1671-0673.2015.05.012
	TONG N , LI L C . Maximum likelihood modulation recognition of MPSK signals based on non-reconstructed compressed samples[J]. Journal of Information Engineering University, 2015, 16 (5): 579- 583. doi: 10.3969/j.issn.1671-0673.2015.05.012
4	马碧云, 元达鹏, 刘娇蛟. 基于似然函数的双曲调频信号参数估计快速算法[J]. 电子与信息学报, 2020, 5, 1228- 1234.
	MA B Y , YUAN D P , LIU J J . A fast algorithm for parameter estimation of hyperbolic frequency modulation signals based on likelihood function[J]. Journal of Electronics and Information Technology, 2020, 5, 1228- 1234.
5	MI X J , TIAN Y , KANG B Y . A hybrid multi-criteria decision making approach for assessing health-care waste management technologies based on soft likelihood function and D-numbers[J]. Applied Intelligence, 2021, 51, 6708- 6727. doi: 10.1007/s10489-020-02148-7
6	郭黎利, 刘湘蒲, 付江志, 等. 基于压缩循环谱的mW复合序列参数估计[J]. 华南理工大学学报(自然科学版), 2016, 44 (5): 29- 35.
	GUO L L , LIU X P , FU J Z , et al. Parameter estimation of MW composite sequence based on compression cycle spectrum[J]. Journal of South China University of Technology (Natural Science), 2016, 44 (5): 29- 35.
7	叶芝慧, 李昂, 彭文攀. 基于混合OFDM调制的认知网络快速组网技术[J]. 东南大学学报(自然科学版), 2017, 47 (4): 637- 641.
	YE Z H , LI A , PENG W P . A fast networking technology of cognitive network based on hybrid OFDM modulation[J]. Journal of Southeast University (Natural Science), 2017, 47 (4): 637- 641.
8	WANG H, GUO L L. A new method of automatic modulation recognition based on dimension reduction[C]//Proc. of the IEEE Forum on Cooperative Positioning and Service, 2017: 316-320.
9	张立民, 凌青, 闫文君. 基于高阶累积量的空时分组码盲识别算法研究[J]. 通信学报, 2016, 37 (5): 1- 8.
	ZHANG L M , LING Q , YAN W J . Research on blind recognition algorithm of space-time block codes based on high-order cumulants[J]. Journal of Communication, 2016, 37 (5): 1- 8.
10	谭晓衡, 褚国星, 张雪静, 等. 基于高阶累积量和小波变换的调制识别算法[J]. 系统工程与电子技术, 2018, 40 (1): 171- 177.
	TANG X H , CHU G X , ZHANG X J , et al. Modulation recogni tion algorithm based on high order cumulant and wavelet transform[J]. Systems Engineering and Electronics, 2018, 40 (1): 171- 177.
11	ABU-ROMOH M , ABOUTALEB A , REZKI Z . Automatic modulation classification using moments and likelihood maximization[J]. IEEE Communications Letters, 2018, 22 (5): 938- 941.
12	O'SHEA T J, CORGAN J, CLANCY T C, et al. Convolutional radio modulation recognition networks[C]//Proc. of the International Conference on Engineering Applications of Neural Networks, 2016: 213-226.
13	WEST N E, O'SHEA T J. Deep architectures for modulation recognition[C]//Proc. of the IEEE International Symposium on Dynamic Spectrum Access Networks, 2017.
14	O'SHEA T J , ROY T , CLANCY T C . Over-the-air deep learning based radio signal classification[J]. IEEE Journal of Selected Topics in Signal Processing, 2018, 12 (1): 168- 179.
15	WANG J , WANG W F , LUO F X , et al. Modulation classification based on denoising autoencoder and convolutional neural network with GNU radio[J]. The Journal of Engineering, 2019, 19, 6188- 6191.
16	FAN M , PENG C , WU L N , et al. Automatic modulation classification: a deep learning enabled approach[J]. IEEE Trans.on Vehicular Technology, 2018, 67 (11): 10760- 10772.
17	QI P H , ZHOU X Y , ZHENG S S , et al. Automatic modulation classification based on deep residual networks with multimodal information[J]. IEEE Trans.on Cognitive Communications and Networking, 2021, 7 (1): 21- 33.
18	ODENA A. Semi-supervised learning with generative adversarial networks[C]//Proc. of the Data Efficient Machine Learning Workshop at International Conference on Machine Learning, 2016.
19	GOODFELLOW I, POUGET-ABADIE J, MIRZA M, et al. Generative adversarial nets[C]//Proc. of the 27th International Conference on Neural Information Processing Systems, 2014, 2: 2672-2680.
20	BU K , HE Y , JING X J , et al. Adversarial transfer learning for deep learning based automatic modulation classification[J]. IEEE Signal Processing Letters, 2020, 27, 880- 884.
21	HE K M, ZHANG X Y, REN S Q, et al. Delving deep into rectifiers: surpassing human-level performance on image net classification[C]//Proc. of the IEEE Conference on Computer Vision and Pattern Recognition, 2016: 1026-1034.
22	HE K M, ZHANG X Y, REN S Q, et al. Deep residual learning for image recognition[C]//Proc. of the IEEE Conference on Computer Vision and Pattern Recognition, 2016: 770-778.

参数	数值
调制样式	8PSK, AM-DSB, BPSK, GFSK CPFSK, PAM4, QAM16, QAM64, QPSK, WBFM
样本维度	(2, 128)
样本数量	86 400
信噪比	-4~18 dB (每2 dB一个间隔)

参数	批次
参数	32	64	128	256	512
第一衰减率	0.4	0.5	0.6	0.5	0.4
Dropout	0.3	0.05	0.15	0.3	0.2
识别准确率	0.918	0.980	0.952	0.934	0.928

网络	参数
网络	网络参数	运行时间/s	分类准确率
CLDNN^[14]	3 616 348	94	0.823
AUCNN^[15]	2 861 583	85	0.887
Resnet-WSMF^[17]	3 522 641	89	0.934
ACGAN^[18]	G: 320 152 D: 268 993	49	0.804
Res-GAN(本文)	G: 310 721 D: 778 291	68	0.980

网络	标签样本量
网络	540	600	660	720	780	1 800	3 600	7 200	15 000	30 000	60 000
CLDNN^[14]	0.192	0.223	0.261	0.285	0.311	0.355	0.441	0.557	0.632	0.774	0.785
AUCNN^[15]	0.221	0.248	0.355	0.374	0.391	0.425	0.504	0.663	0.724	0.827	0.832
Resnet-WSMF^[17]	0.191	0.198	0.254	0.284	0.316	0.345	0.397	0.445	0.664	0.832	0.895
ACGAN^[18]	0.233	0.243	0.265	0.287	0.328	0.391	0.462	0.683	0.764	0.825	0.863
Res-GAN	0.849	0.897	0.980	0.979	0.978	0.979	0.975	0.980	0.979	0.977	0.978

[1]	康颖, 赵治华, 吴灏, 李亚星, 孟进. 基于Deep SVDD的通信信号异常检测方法[J]. 系统工程与电子技术, 2022, 44(7): 2319-2328.
[2]	孙晶明, 虞盛康, 孙俊. 基于元学习的雷达小样本目标识别方法及改进[J]. 系统工程与电子技术, 2022, 44(6): 1839-1845.
[3]	侯召国, 王华伟, 周良, 付强. 基于改进深度残差网络的旋转机械故障诊断[J]. 系统工程与电子技术, 2022, 44(6): 2051-2059.
[4]	邵凯, 朱苗苗, 王光宇. 基于生成对抗与卷积神经网络的调制识别方法[J]. 系统工程与电子技术, 2022, 44(3): 1036-1043.
[5]	吕勤哲, 全英汇, 沙明辉, 董淑仙, 邢孟道. 基于集成深度学习的有源干扰智能分类[J]. 系统工程与电子技术, 2022, 44(12): 3595-3602.
[6]	徐桂光, 王旭东, 汪飞, 胡国兵, 高涌荇, 罗泽虎. 基于SE-ResNeXt网络的低信噪比LPI雷达辐射源信号识别[J]. 系统工程与电子技术, 2022, 44(12): 3676-3684.
[7]	刘旗, 张新禹, 刘永祥. 基于门控多尺度匹配网络的小样本SAR目标识别[J]. 系统工程与电子技术, 2022, 44(11): 3346-3356.
[8]	马玉凤, 向南, 豆亚杰, 姜江, 杨克巍, 谭跃进. 军事系统工程中的知识图谱应用及研究[J]. 系统工程与电子技术, 2022, 44(1): 146-153.
[9]	史蕴豪, 许华, 郑万泽, 刘英辉. 基于集成学习与特征降维的小样本调制识别方法[J]. 系统工程与电子技术, 2021, 43(4): 1099-1109.
[10]	晏媛, 孙俊, 孙晶明, 于俊朋. 雷达小样本目标识别方法及应用分析[J]. 系统工程与电子技术, 2021, 43(3): 684-692.
[11]	张良, 杨威, 李玮杰, 杨小琪, 刘永祥. 基于复数卷积-残差网络的雷达杂波幅度统计模型分类[J]. 系统工程与电子技术, 2021, 43(11): 3086-3097.
[12]	刘凯, 张斌, 黄青华. 基于TCNN-BiLSTM网络的调制识别算法[J]. 系统工程与电子技术, 2020, 42(8): 1841-1849.
[13]	曾强, 黄政, 魏曙寰. 融合专家先验知识和单调性约束的贝叶斯网络参数学习方法[J]. 系统工程与电子技术, 2020, 42(3): 646-652.
[14]	刘凯, 赵梦伟, 黄青华. 基于近似熵的Multi-h CPM调制识别算法[J]. 系统工程与电子技术, 2020, 42(3): 698-703.
[15]	石礼盟, 杨承志, 吴宏超. 基于深层残差网络和三元组损失的雷达信号识别方法[J]. 系统工程与电子技术, 2020, 42(11): 2506-2512.