Automatic Modulation Recognition Algorithm Based on Multi-Channel Joint Learning

doi:10.3969/j.issn.1671-1122.2023.04.003

Abstract

Abstract:

Automatic modulation recognition technology can not only effectively improve the utilization rate of spectrum resources, but is also an effective way to identify illegal users. To further improve the performance of the recognition algorithm, the paper proposed a new asymmetric multichannel joint learning network by considering the connection between amplitude and phase features. The network used the amplitude, phase and the joint matrix of both as multi-channel input to achieve adaptive modulation coding by better extracting homogeneous and heterogeneous features in the amplitude and phase of the modulated signal using an asymmetric joint learning module without changing the number of parameters and computational speed. The experiments results show that the network proposed in the article achieves the highest recognition accuracy of 91.73% and 93.36% on the benchmark open source datasets RadioML2016.10a and RadioML2016.10b, respectively.

Key words: deep learning, modulation recognition, convolutional neural networks, joint learning

CLC Number:

TP309

ZHAO Caidan, CHEN Jingqian, WU Zhiqiang. Automatic Modulation Recognition Algorithm Based on Multi-Channel Joint Learning[J]. Netinfo Security, 2023, 23(4): 20-29.

Figures/Tables 9

References 23

[1]	GUPTA A, JHA R K. A Survey of 5G Network: Architecture and Emerging Technologies[J]. IEEE Access, 2015, 3: 1206-1232. doi: 10.1109/ACCESS.2015.2461602 URL
[2]	WANG Qun, SUN Haijian, HU R Q Y, et al. When Machine Learning Meets Spectrum Sharing Security: Methodologies and Challenges[J]. IEEE Open Journal of the Communications Society, 2022, 3: 176-208. doi: 10.1109/OJCOMS.2022.3146364 URL
[3]	LANZOLLA G, ANDERSON J A. Digital Transformation[J]. Business Strategy Review, 2008, 19(2): 72-76. doi: 10.1111/j.1467-8616.2008.00539.x URL
[4]	ZHANG Lin, XIAO Ming, WU Gang, et al. A Survey of Advanced Techniques for Spectrum Sharing in 5G Networks[J]. IEEE Wireless Communications, 2017, 24(5): 44-51. doi: 10.1109/MWC.2017.1700069 URL
[5]	ALARBI A, ALKISHRIWO O A S. Modulation Classification Based on Statistical Features and Artificial Neural Network[C]// IEEE. 2021 IEEE 1st International Maghreb Meeting of the Conference on Sciences and Techniques of Automatic Control and Computer Engineering MI-STA. New York: IEEE, 2021: 748-751.
[6]	KIM K, POLYDOROS A. Digital Modulation Classification: The BPSK Versus QPSK Case[C]// IEEE. MILCOM 88, 21st Century Military Communications Conference. New York: IEEE, 1988: 431-436.
[7]	JIANG Yuhan, ZOU Yulong, OUYANG Jian, et al. Secrecy Energy Efficiency Optimization for Artificial Noise Aided Physical-Layer Security in OFDM-Based Cognitive Radio Networks[J]. IEEE Transactions on Vehicular Technology, 2018, 67(12): 11858-11872. doi: 10.1109/TVT.2018.2876062 URL
[8]	LI Xiaofan, DONG Fangwei, ZHANG Sha, et al. A Survey on Deep Learning Techniques in Wireless Signal Recognition[J]. Wireless Communications and Mobile Computing, 2019, 179: 1-12.
[9]	WALENCZYKOWSKA M, KAWALEC A. Type of Modulation Identification Using Wavelet Transform and Neural Network[J]. Bulletin of the Polish Academy of Sciences Technical Sciences, 2016, 64(1): 257-261. doi: 10.1515/bpasts-2016-0028 URL
[10]	SWAMI A, SADLER B M. Hierarchical Digital Modulation Classification Using Cumulants[J]. IEEE Transactions on Communications, 2000, 48(3): 416-429. doi: 10.1109/26.837045 URL
[11]	VUCIC D, VUKOTIC S, ERIC M. Cyclic Spectral Analysis of OFDM/OQAM Signals[J]. AEU-International Journal of Electronics and Communications, 2017, 73: 139-143.
[12]	OTTER D W, MEDINA J R, KALITA J K. A Survey of the Usages of Deep Learning for Natural Language Processing[J]. IEEE Transactions on Neural Networks and Learning Systems, 2021, 32(2): 604-624. doi: 10.1109/TNNLS.5962385 URL
[13]	KHAN M, SAJJAD M, KWON S. Clustering-Based Speech Emotion Recognition by Incorporating Learned Features and Deep BiLSTM[J]. IEEE Access, 2020, 8: 79861-79875. doi: 10.1109/Access.6287639 URL
[14]	O’SHEA T J, CORGAN J, CLANCY T C. Convolutional Radio Modulation Recognition Networks[C]// Springer. International Conference on Engineering Applications of Neural Networks. Berlin:Springer, 2016: 213-226.
[15]	WEST N E, O'SHEA T. Deep Architectures for Modulation Recognition[C]// IEEE. 2017 IEEE International Symposium on Dynamic Spectrum Access Networks (DySPAN). New York: IEEE, 2017: 1-7.
[16]	RAJENDRAN S, MEERT W, GIUSTINIANO D, et al. Deep Learning Models for Wireless Signal Classification with Distributed Low-Cost Spectrum Sensors[J]. IEEE Transactions on Cognitive Communications and Networking, 2018, 4(3): 433-445. doi: 10.1109/TCCN.2018.2835460 URL
[17]	PERENDA E, RAJENDRAN S, POLLIN S. Automatic Modulation Classification Using Parallel Fusion of Convolutional Neural Networks[C]// BalkanCom. Third International Balkan Conference on Communications and Networking. Istanbul: BalkanCom, 2019: 469-478.
[18]	XU Jialang, LUO Chunbo, PARR G, et al. A Spatiotemporal Multi-Channel Learning Framework for Automatic Modulation Recognition[J]. IEEE Wireless Communications Letters, 2020, 9(10): 1629-1632. doi: 10.1109/LWC.5962382 URL
[19]	LIAO Kaisheng, ZHAO Yaodong, GU Jie, et al. Sequential Convolutional Recurrent Neural Networks for Fast Automatic Modulation Classification[J]. IEEE Access, 2021, 9: 27182-27188. doi: 10.1109/Access.6287639 URL
[20]	HUYNH-THE T, PHAM Q V, NGUYEN T V, et al. Automatic Modulation Classification: A Deep Architecture Survey[J]. IEEE Access, 2021, 9: 142950-142971. doi: 10.1109/ACCESS.2021.3120419 URL
[21]	LIU Minhao, ZENG Ailing, LAI Q, et al. Time Series is a Special Sequence: Forecasting with Sample Convolution and Interaction[EB/OL]. (2021-10-19)[2022-11-05]. https://www.xueshufan.com/publication/3171665133.
[22]	O'SHEA T J, WEST N. Radio Machine Learning Dataset Generation with GNU Radio[EB/OL]. (2016-09-06)[2022-11-05]. https://pubs.gnuradio.org/index.php/grcon/article/view/11.
[23]	LAURENS V D M, GEOFFREY H. Visualizing Data Using t-SNE[J]. Journal of Machine Learning Research, 2008, 9(11): 2579-2625.

组成	描述	输出大小/张量
输入	AP 数值	2, 128
symmetric/ asymmetric-MJLNet		2, 128, 64
一维卷积Conv1	conv(64, 2*8, 1)	1, 128, 64
级联层Concat	Concat()	2, 128, 128
全局平均池化层AVGPOOL	GlobleAveragePooling2D()	1, 128
长短期记忆层LSTM * 2	LSTM(128) × 2	128
输出	Fc(11/10)	10/11

网络调制方式	SCRNN	1DCNN-PF	CNN2	CLDNN	MCLDNN	LSTM2	SCRNN	asymmetric-MJLNet
8PSK	84%	80%	77%	82%	86%	73%	84%	89%
AM-DSB	86%	99%	99%	87%	89%	93%	86%	90%
AM-SSB	97%	98%	95%	94%	92%	95%	97%	93%
BPSK	99%	98%	98%	99%	99%	94%	99%	99%
CPFSK	100%	100%	98%	96%	100%	99%	100%	100%
GFSK	99%	94%	94%	93%	99%	97%	99%	97%
PAM4	98%	98%	98%	97%	98%	96%	98%	98%
QAM16	52%	83%	40%	40%	81%	76%	52%	83%
QAM64	55%	71%	59%	63%	86%	48%	55%	81%
QPSK	82%	95%	40%	74%	96%	86%	82%	97%
WBFM	39%	29%	24%	43%	41%	33%	39%	38%

数据集	网络	识别率	训练时间/s	参数量/k	FLOPs/k
RadioML 2016.10a	CNN2	53.40%	17	2750	4648
	1DCNN-PF	57.63%	18	179	7514
	CLDNN	57.03%	14	164	6450
	MCLDNN	61.50%	31	406	1964
	LSTM2	56.79%	20	201	6082
	SCRNN	57.21%	11	40	5308
	asymmetric-MJLNet	61.70%	25	530	7238