基于RE-GAN的调制信号开集识别算法

doi:10.12305/j.issn.1001-506X.2023.10.37

摘要/Abstract

摘要：

为解决轻量化模型下调制信号开集识别准确率不高的问题, 设计了一种数据重建和极值理论生成对抗网络模型, 包含一对相互对抗的网络: 重建网络和判别网络。首先, 重建网络利用自编码器对信号进行压缩重建, 将高维度数据压缩为低维表达。然后, 判别网络对压缩后的数据进行特征提取并建立各类数据的激活矢量集用以拟合极值分布。最后, 通过极值理论求出已知和未知调制方式信号的概率。仿真实验表明, 该模型显著降低了算法复杂度, 不仅能对已知调制方式的信号进行充分学习和表达, 还能干扰未知调制方式的信号, 在信噪比大于0 dB时, 针对8种已知调制方式和2种未知调制方式信号的识别准确率均达到93%。

关键词: 数据重建, 极值理论, 生成对抗网络, 开集识别, 激活矢量

Abstract:

In order to solve the problem of open-set low recognition accuracy of modulation signal under the lightweight model, a model of data reconstruction and extremum theory generative adversarial network is designed. This model includes a pair of adversarial networks: reconstructed network and discriminant network. Firstly, the reconstructed network uses the autoencoder to compress and reconstruct the signal, and compresses the high-dimensional data into low-dimensional representation. Then, the discriminant network extracts the features of the compressed data and establishes the activation vector set of all kinds of data to fit the extremum distribution. Finally, the probability of known and unknown modulation signals is calculated by extremum theory. The experiment results show that the proposed model can not only fully learn and express the known modulation signals, but also disturb the unknown modulation signals. When the signal to noise ratio is greater than 0 dB, the recognition accuracy of eight known modulation signals and two unknown modulation signals can reach 93%.

Key words: data reconstruction, extreme value theory, generative adversarial network (GAN), open-set recognition, activation vector

中图分类号:

TN975

秦博伟, 蒋磊, 许华, 牛伟宇. 基于RE-GAN的调制信号开集识别算法[J]. 系统工程与电子技术, 2023, 45(10): 3321-3328.

Bowei QIN, Lei JIANG, Hua XU, Weiyu NIU. Open-set recognition algorithm for modulation signal based on RE-GAN[J]. Systems Engineering and Electronics, 2023, 45(10): 3321-3328.

图/表 16

图1

图2

图3

图4

表1

图5

图6

图7

图8

图9

图10

图11

图12

图13

图14

表2

参考文献 14

15	PATEL M, WANG X Y, MAO S W. Data augmentation with conditional GAN for automatic modulation classification[C]//Proc. of the 2nd ACM Workshop on Wireless Security and Machine Learning, 2020.
16	GOODFELLOW I, POUGET-ABADIE J, MIRZA M, et al. Generative adversarial networks[C]//Proc. of the Conference and Workshop on Neural Information Processing Systems, 2014: 2672-2680.
17	BENDALE A, BOULT T. Towards open-set deep networks[C]//Proc. of the IEEE Conference on Computer Vision and Pattern Recognition, 2016: 1563-1572.
18	GE Z Y, DEMYANOV S, CHEN Z T, et al. Generative openmax for multi-class open set classification[C]//Proc. of the British Machine Vision Conference, 2017.
19	NEAL L, OLSON M, FERN X, et al. Open-set learning with counterfactual images[C]//Proc. of the European Conference on Computer Vision, 2018.
20	RUDD E , JAIN L P , SCHEIRER W J , et al. The extreme va-lue machine[J]. IEEE Trans.on Pattern Analysis and Machine Intelligence, 2018, 40 (3): 762- 768.
21	LUCIC M, KURACH K, MICHALSKI M, et al. Are GANs created equal? A large-scale study[C]//Proc. of the Conference and Workshop on Neural Information Processing Systems, 2018.
22	TANG B , TU Y , ZHANG Z 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.
23	YOSHIHASHI R, SHAO W, KAWAKAMI R, et al. Classification-reconstruction learning for open-set recognition[C]//Proc. of the IEEE Conference on Computer Vision and Pattern Recognition, 2019.
1	郝云飞, 刘章孟, 郭福成, 等. 基于生成对抗网络的信号调制方式的开集识别[J]. 系统工程与电子技术, 2019, 41 (11): 2619- 2624.
	HAO Y F , LIU Z M , GUO F C , et al. Open-set recognition of signal modulation based on generative adversarial networks[J]. Systems Engineering and Electronics, 2019, 41 (11): 2619- 2624.
2	HAN Y S , WU T Y , CHEN P N , et al. A low-complexity maximum-likelihood decoder for tail-biting convolutional codes[J]. IEEE Trans.on Communications, 2018, 66 (5): 1859- 1870. doi: 10.1109/TCOMM.2018.2790935
3	CHO B J , LEE J M , PARK H M . A beam-forming algorithm based on maximum likelihood of a complex Gaussian distribution with time-varying variances for robust speech recognition[J]. IEEE Signal Processing Letters, 2019, 26 (9): 1398- 1402. doi: 10.1109/LSP.2019.2932848
4	胡国兵, 吴珊珊, 杨忠, 等. LFM/BPSK混合调制信号盲处理结果可信性评估: 一种简化的似然比算法[J]. 电子学报, 2019, 47 (9): 1891- 1897. doi: 10.3969/j.issn.0372-2112.2019.09.011
	HU G B , WU S S , YANG Z , et al. Credibility evaluation of blind processing results of LFM/BPSK mixed modulation signals: a simplified likelihood ratio algorithm[J]. Acta Electronica Sinica, 2019, 47 (9): 1891- 1897. doi: 10.3969/j.issn.0372-2112.2019.09.011
5	ABDELBAR M , TRANTER W H , BOSE T . Cooperative cumulants-based modulation classification in distributed networks[J]. IEEE Trans.on Cognitive Communications and Networking, 2018, 4 (3): 446- 461. doi: 10.1109/TCCN.2018.2824326
6	李红光, 郭英, 眭萍, 等. 基于时频能量谱纹理特征的跳频调制方式识别[J]. 通信学报, 2019, 40 (10): 20- 29.
	LI H G , GUO Y , SUI P , et al. Recognition of frequency hopping modulation based on texture features of time-frequency energy spectrum[J]. Journal of Communications, 2019, 40 (10): 20- 29.
7	张天骐, 范聪聪, 葛宛营, 等. 基于ICA和特征提取的MIMO信号调制识别算法[J]. 电子与信息学报, 2020, 42 (9): 2208- 2215.
	ZHANG T Q , FAN C C , GE W Y , et al. Modulation recognition algorithm of MIMO signal based on ICA and feature extraction[J]. Journal of Electronics and Information, 2020, 42 (9): 2208- 2215.
8	RAJENDRAN S , MEERT W , GIUSTINIANO D , et al. Deep learning models for wireless signal classification with distributed low-cost spectrum sensors[J]. IEEE Trans.on Cognitive Communications and Networking, 2018, 4 (3): 433- 445. doi: 10.1109/TCCN.2018.2835460
9	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.
10	张思成, 林云, 涂涯, 等. 基于轻量级深度神经网络的电磁信号调制识别技术[J]. 通信学报, 2020, 41 (11): 12- 21.
	ZHANG S C , LIN Y , TU Y , et al. Electromagnetic signal modulation recognition technology based on lightweight deep neural network[J]. Journal of Communications, 2020, 41 (11): 12- 21.
11	安泽亮, 张天骐, 马宝泽, 等. 基于一维CNN的多入多出OSTBC信号协作调制识别[J]. 通信学报, 2021, 42 (7): 84- 94.
	AN Z L , ZHANG T Q , MA B Z , et al. Cooperative modulation recognition of multiple-input multiple-output OSTBC signal based on one-dimensional CNN[J]. Journal of Communications, 2021, 42 (7): 84- 94.
12	肖易寒, 王亮, 郭玉霞. 基于去噪卷积神经网络的雷达信号调制类型识别[J]. 电子与信息学报, 2021, 43 (8): 2300- 2307.
	XIAO Y H , WANG L , GUO Y X . Modulation type recognition of radar signal based on denoising convolutional neural network[J]. Journal of Electronics and Information, 2021, 43 (8): 2300- 2307.
13	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.
14	LI M X, LIU G Y, LI S T, et al. Radio classify generative adversarial networks: a semi-supervised method for modulation recognition[C]//Proc. of the IEEE 18th International Confe-rence on Communication Technology, 2018: 669-672.

参数	取值
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	训练: 115 200 测试: 36 000
SNR	-4~18 dB (每2 dB一个间隔)

算法	网络参数	模型乘/加次数	运行时间/h	闭集识别平均分类准确率	开集识别平均分类准确率
Softmax	G: 866 88	G: 172 045	训练: 1.67	0.941	-
Softmax	D: 260 837	D: 530 118	测试: 0.03	(±0.63)	-
Openmax^[18]	654 527	1 276 481	训练: 1.20	0.803	0.845
Openmax^[18]	654 527	1 276 481	测试: 0.02	(±0.92)	(±0.65)
G-Openmax^[19]	G: 112 647	G: 228 724	训练: 1.93	0.887	0.896
G-Openmax^[19]	D: 334 658	D: 667 852	测试: 0.07	(±0.74)	(±0.37)
OLTR^[23]	867 439	1 674 892	训练: 2.18	0.894	0.905
OLTR^[23]	867 439	1 674 892	测试: 0.06	(±0.43)	(±0.77)
RE-GAN	G: 86 688	G: 172 045	训练: 1.80	0.938	0.934
RE-GAN	D: 280 841	D: 560 722	测试: 0.04	(±0.71)	(±0.68)

[1]	庞世燕, 郝京京, 邢立宁, 谭旭. 融合域自适应和弱监督策略的遥感影像建筑物提取算法[J]. 系统工程与电子技术, 2023, 45(6): 1616-1623.
[2]	陈丽, 方梓涵, 梅立泉. 基于GAN的直扩信号生成算法[J]. 系统工程与电子技术, 2023, 45(5): 1544-1552.
[3]	曾浩然, 冯蕴雯, 路成, 潘维煌. 基于改进生成对抗网络的国产民机航材消耗预测方法[J]. 系统工程与电子技术, 2023, 45(10): 3132-3138.
[4]	韩啸, 陈世文, 陈蒙, 杨锦程. 基于互易点学习的LPI信号开集识别[J]. 系统工程与电子技术, 2022, 44(9): 2752-2759.
[5]	刘戎翔, 吴琳, 谢智歌, 刘虹麟. 基于生成对抗网络的防空体系态势辅助分析[J]. 系统工程与电子技术, 2022, 44(8): 2522-2529.
[6]	徐平亮, 崔亚奇, 熊伟, 熊振宇, 顾祥岐. 生成式中断航迹接续关联方法[J]. 系统工程与电子技术, 2022, 44(5): 1543-1552.
[7]	曹鹏宇, 杨承志, 石礼盟, 吴宏超. 基于PSO-DBSCAN和SCGAN的未知雷达信号处理方法[J]. 系统工程与电子技术, 2022, 44(4): 1158-1165.
[8]	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.
[9]	曹鹏宇, 杨承志, 石礼盟, 吴宏超. 基于DAE-GAN网络的LPI雷达信号增强[J]. 系统工程与电子技术, 2021, 43(9): 2493-2500.
[10]	丁嘉辉, 汤建龙, 于正洋. 轻量化的增量式集成学习算法设计[J]. 系统工程与电子技术, 2021, 43(4): 861-867.
[11]	赵凡, 金虎. 基于GAN的通信干扰波形生成技术[J]. 系统工程与电子技术, 2021, 43(4): 1080-1088.
[12]	田相轩, 石志强. 基于改进型生成对抗网络的指挥信息系统模拟数据生成算法[J]. 系统工程与电子技术, 2021, 43(1): 163-170.
[13]	向南, 张雄涛, 豆亚杰, 徐向前, 杨克巍, 谭跃进. 基于生成对抗网络的高端装备研制数据脱敏方法[J]. 系统工程与电子技术, 2020, 42(6): 1310-1316.
[14]	殷小静, 胡晓峰, 刘戎翔, 谢智歌, 马骏. 基于GAN的体系能力图谱生成方法[J]. 系统工程与电子技术, 2020, 42(10): 2257-2264.
[15]	朱克凡, 王杰贵. 小样本条件下SCGAN+CNN低分辨雷达目标一步识别算法[J]. 系统工程与电子技术, 2020, 42(1): 67-75.