无线电信号的识别技术是保障电磁空间安全的重要基础，然而，传统的非合作通信信号识别方法难以应对无线通信新业务需求和信号复杂多变的情况。深度学习在挖掘数据特征方面具有显著优势，被广泛应用于电磁信号数据的处理。因此，本文针对非合作通信信号的识别问题，采用深度学习算法，从调制-编码联合识别、通信体制识别和频谱识别三个方面展开研究，对智能频谱管控和电子侦察对抗具有重要意义。

论文主要工作有以下几个方面：

(1)针对非合作通信中编码方法和调制方式不能同时识别的问题，提出了基于深度学习的信号调制-编码联合识别方法。构建了由八种信道编码和五种调制方式组成的数据集，设计了六种不同的网络模型，从不同数据预处理方法、样本量、信道条件、网络模型、天线数目和脉冲噪声等方面进行仿真实验，结果表明基于深度学习的非合作通信信号联合识别方法能够以较高的识别准确率同时识别信号的编码类型与调制方式。

(2)针对通用移动通信系统（Universal Mobile Telecommunications System，UMTS）、长期演进（Long-Term Evolution，LTE）和第五代新无线电（5th Generation New Radio，5G NR）这三种蜂窝通信体制的识别问题，从信号角度出发，提出了基于深度学习的蜂窝通信体制识别方法。利用MATLAB仿真生成蜂窝信号数据集，对不同网络模型进行优化训练，着重考虑了不同输入模态、特征量、信道条件、网络模型和脉冲噪声对蜂窝通信体制识别性能的影响，结果表明基于深度学习的蜂窝通信体制识别方法在较高信噪比下可以准确识别不同的蜂窝通信体制。

(3)针对LTE和5G NR信号的频谱识别问题，提出了基于语义分割网络的LTE与5G NR信号频谱识别方法。利用MATLAB仿真生成LTE和5G NR信号的频谱图，采用语义分割网络和迁移学习技术对LTE和5G NR信号的频谱进行识别。从不同频谱图大小、迁移学习模型、信噪比和混合频谱图等方面进行仿真实验，结果表明基于语义分割的LTE与5G NR信号频谱识别方法在较高信噪比下识别准确率达到了98%以上。

The technology of radio signal recognition serves as a crucial foundation for ensuring electromagnetic space security. However, traditional non-cooperative communication signal recognition methods struggle to meet the demands of new wireless communication services and the complexity of signals. Deep learning offers significant advantages in extracting data features and has been widely applied to the processing of electromagnetic signal data. Therefore, this thesis addresses the problem of identifying non-cooperative communication signals using deep learning algorithms. It focuses on three aspects: joint recognition of modulation and encoding, recognition of communication regimes, and spectrum recognition, which are of great importance for intelligent spectrum management and electronic reconnaissance countermeasures.

The main work of this thesis includes the following aspects:

(1)Addressing the issue that encoding methods and modulation modes in non-cooperative communication cannot be identified simultaneously, a deep learning-based joint recognition method for signal modulation and encoding is proposed. A dataset comprising eight channel encoding types and five modulation methods was constructed. Six different neural network models were designed, and simulation experiments were conducted from various perspectives, including data preprocessing, sample size, channel conditions, network models, number of antennas, and pulse noise. The results demonstrate that the deep learning-based joint recognition method for non-cooperative communication signals can identify the signal's encoding type and modulation mode simultaneously with high accuracy.

(2)Addressing the challenge of identifying different cellular communication systems, such as Universal Mobile Telecommunications System (UMTS), Long-Term Evolution (LTE), and 5th Generation New Radio (5G NR), a deep learning-based method for cellular communication system recognition from a signal perspective is proposed. Cellular signal datasets were generated using MATLAB simulations for training and optimizing various network models. The research focuses on the effects of different input modalities, feature quantities, channel conditions, network models, and pulse noise on the method for identifying cellular communication systems. The results demonstrate that the deep learning-based method can accurately identify different cellular communication regimes at higher signal-to-noise ratios.

(3)Addressing the spectral recognition challenges of LTE and 5G NR signals, this thesis introduces a method based on semantic segmentation networks for the recognition of LTE and 5G NR signal spectra. Using MATLAB simulations, spectrograms of LTE and 5G NR signals are generated, which are then analyzed using semantic segmentation networks coupled with transfer learning techniques. The experiments are conducted with variations in spectrogram sizes, transfer learning models, signal-to-noise ratios, and mixed spectra. The results demonstrate that the semantic segmentation-based approach achieves an accuracy of over 98% in recognizing the spectra of LTE and 5G NR signals at high signal-to-noise ratios.

[1]宋叶辉, 丁国如, 徐承龙, 孙佳琛, 汤鹏. 面向非合作无人机通信网络的通联拓扑推理技术[J]. 电子与信息学报, 2022, 44(3): 924-939.

[2]Bande M, Magesh A, Veeravalli V V. Dynamic spectrum access using stochastic multi-user bandits[J]. IEEE Wireless Communications Letters, 2021, 10(5): 953-956.

[3]李鹏程, 顾杰, 王淑君, 等. 基于第三代移动通信CDMA基站信号的非合作定位技术[J]. 电子信息对抗技术, 2022, 37(1): 5-16.

[4]薛德鑫, 单涛, 董士军, 张进. 一种非协作通信信号调制方式识别方法[J]. 现代防御技术, 2022, 50(5): 152-159.

[5]Bai L, Huang Z, Feng T, et al. A non-stationary channel model for 6G multi-UAV cooperative communication[J]. IEEE Transactions on Wireless Communications, 2023, 13(7): 76-85.

[6]Liu M, Zhang Z, Chen Y, et al. Adversarial attack and defense on deep learning for air transportation communication jamming[J]. IEEE Transactions on Intelligent Transportation Systems, 2023, 13(4): 1-13.

[7]Ramabadran S, Madhukumar A S. Blind parameter estimation of Turbo convolutional codes: noisy and non-synchronized scenario[J]. Digital Signal Processing, 2019, 95: 102577.

[8]Kumar, Madhu A S, Guohua, et al. Blind recognition of LDPC code parameters over erroneous channel conditions[J]. IET Signal Processing, 2019, 13(1): 86-95.

[9]Wu G, Zhang B, Gu D, et al. Blind recognition of BCH codes in faster-than-nyquist signalling system[J]. Electronics Letters, 2016, 52(9): 716-718.

[10]Giard P, Balatsoukas-Stimming A, Burg A. Blind detection of polar codes[C]//2017 IEEE International Workshop on Signal Processing Systems (SiPS). IEEE, 2017, 3(1): 1-6.

[11]Swaminathan R, Madhukumar A S. Classification of error correcting codes and estimation of interleaver parameters in a noisy transmission environment[J]. IEEE Transactions on Broadcasting, 2018, 63(3): 463-478.

[12]东阳. Turbo码盲识别技术研究与实现[D]. 电子科技大学, 2024, 13(7): 1-15.

[13]Yu P, Li J, Peng H. Gibbs sampling based parameter estimation for RSC subcodes of turbo codes[C]//International Conference on Wireless Communications & Signal Processing. IEEE, 2014, 34(1): 1-15.

[14]张天骐, 王雪怡, 方竹, 等. 基于软判决下的校验关系的循环码参数盲识别[J]. 信号处理, 2023, 39(6): 1120-1130.

[15]付卫红, 刘乃安, 郑靖轩. 卷积码盲识别树图遍历算法研究与仿真实验[J]. 现代电子技术, 2023, 46(7): 5-9.

[16]陶志勇, 闫明豪, 刘影. 基于时序卷积网络的信道编码闭集识别[J]. 华中科技大学学报: 自然科学版, 2022, 50(3): 6.

[17]刘洁, 刘凯. 基于深度学习的纠错码方式识别[J]. 电子测量技术, 2019, 16(8): 154-158.

[18]Qin X, Peng S, Yang X, et al. Deep learning based channel code recognition using TextCNN[C]//2019 IEEE International Symposium on Dynamic Spectrum Access Networks (DySPAN). IEEE, 2019, 44(1): 1-5.

[19]Dehdashtian S, Hashemi M, Salehkaleybar S. Deep-learning-based blind recognition of channel code parameters over candidate sets under AWGN and multi-path fading conditions[J]. IEEE Wireless Communication Letters, 2021, 99: 1-1.

[20]Peng S, Jiang H, Wang H, et al. Modulation classification based on signal constellation diagrams and deep learning[J]. IEEE Transactions on Neural Networks and Learning Systems, 2018, 99: 1-10.

[21]Zhang H, Nie R, Lin M, et al. A deep learning based algorithm with multi-level feature extraction for automatic modulation recognition[J]. Wireless Networks, 2021, 27(7): 4665-4676.

[22]Zhang Z, Liu F, Sheng J, Huang C and Shi B, Index modulation recognition based on projection residual analysis[J]. IEEE Transactions on Cognitive Communications and Networking, 2022, 8(3): 1387-1398.

[23]Liu K, Xiang X, Liang Y, et al. Automatic modulation recognition through wireless sensor networks in aeronautical wireless channel[J]. IEEE Sensors Journal, 2021, 21(20): 23125-23132.

[24]Lin S, Zeng Y, Gong Y. Learning of time-frequency attention mechanism for automatic modulation recognition[J]. IEEE Wireless Communications Letters, 2022, 11(4): 707-711.

[25]Chen B Y, Shen Y K, Sun K. Research on object detection algorithm based on multilayer information fusion[J]. Mathematical Problems in Engineering, 2020, 2020 27(7): 1-13.

[26]An Z, Zhang T, Ma B,et al. A two-stage high-order modulation recognition based on projected accumulated constellation vector in non-cooperative B5G OSTBC-OFDM systems[J]. Signal Processing: The Official Publication of the European Association for Signal Processing (EURASIP), 2022 27(7): 1-15.

[27]Karami E, Dobre O A, Adnani N. Identification of GSM and LTE signals using their second-order cyclostationarity[C]//2015 IEEE International Instrumentation and Measurement Technology Conference (I2MTC) Proceedings. IEEE, 2015 2(7): 1108-1112.

[28]Eldemerdash Y A, Dobre O A, Üreten O, et al. Fast and robust identification of GSM and LTE signals[C]//2017 IEEE International Instrumentation and Measurement Technology Conference (I2MTC). IEEE, 2017 57(7): 1-6.

[29]Eldemerdash Y A, Dobre O A,Oktay Üreten, et al. Identification of cellular networks for intelligent radio measurements[J]. IEEE Transactions on Instrumentation & Measurement, 2017, 8: 1-8.

[30]Alyaoui N, Kachouri A, Samet M. The fourth generation 3GPP LTE identification for cognitive radio[C]//International Conference on Microelectronics. IEEE, 2011 27(6): 1-15.

[31]Abrishamkar F, Butler B, Mohseni M, et al. Method and apparatus for pilot estimation using suboptimum expectation maximization: U.S. Patent Application 10/262,306[P]. 2003-12-11.

[32]Oner M, Jondral F. Cyclostationarity based air interface recognition for software radio systems[C]//Proceedings. 2004 IEEE Radio and Wireless Conference. IEEE, 2004: 263-266.

[33]Alshathri K, Xia H, Lawrence V, et al. Cellular system identification using deep learning: GSM, UMTS and LTE[C]//2019 28th Wireless and Optical Communications Conference (WOCC), 2019: 345-361.

[34]Xia H, Alshathri K, Lawrence V. B, Yao Y.-D, Montalvo A, Rauchwerk M, et al. Cellular signal identification using convolutional neural networks: AWGN and rayleigh fading channels[C]//2019 IEEE International Symposium on Dyndpmic Spectrum Access Networks (DySPAN). IEEE, 2019: 1-5.

[35]杨保平, 陈永光, 谢贵武. 频谱混叠无线电信号盲侦察识别技术[J]. 系统仿真学报, 2014, 8: 7.

[36]陈星, 贺志强, 吴伟陵. 基于循环平稳的多天线感知无线电频谱检测[J]. 北京邮电大学学报, 2008, 2: 89-93.

[37]Selim A, Paisana F, Arokkiam J A, et al. Spectrum monitoring for radar bands using deep convolutional neural networks[C]//GLOBECOM 2017-2017 IEEE Global Communications Conference. IEEE, 2017: 1-6.

[38]周宇航, 侯进, 李嘉新, 等. 基于频域叠加和深度学习的频谱信号识别[J]. 计算机应用研究, 2023, 40(3): 6.

[39]Aghabeiki S, Hallet C, Noutehou N E R, et al. Machine-learning-based spectrum sensing enhancement for software-defined radio applications[C]//2021 IEEE Cognitive Communications for Aerospace Applications Workshop (CCAAW). IEEE, 2021: 1-6.

[40]Ahmed R, Chen Y, Hassan B. Deep residual learning-based cognitive model for detection and classification of transmitted signal patterns in 5G smart city networks [J]. Digital Signal Processing, 2022(120): 120.

[41]Challita U, Sandberg D. Deep reinforcement learning for dynamic spectrum sharing of LTE and NR[C]//ICC 2021-IEEE International Conference on Communications. IEEE, 2021: 1-6.

[42]殷赞, 王超杰, 程子恒, 等. 一种基于注意力机制卷积神经网络模型的自动调制识别算法[J].电波科学学报, 2023, 5:773-779.

[43]Song L, Sheng W, et al. Research on radar clutter recognition method based on LSTM[J]. The Journal of Engineering, 2019, 19: 6247-6251.

[44]Behura S, Kedia S, Hiremath S M, et al. WiST ID-Deep learning-based large scale wireless standard technology identification[J]. IEEE Transactions on Cognitive Communications and Networking, 2020, 6(4): 1365-1377.

[45]Kürat Tekbyk, Akbunar Z, Ekti A R, et al. Spectrum sensing and signal identification with deep learning based on spectral correlation function[J]. IEEE Transactions on Vehicular Technology, 2021, 10: 70.

[46]Gao Y, Ma S, Shi J, et al. An optimised multi-channel neural network model based on ResLDNN for automatic modulation recognition[J]. International Journal of Wireless and Mobile Computing, 2023, 24(2): 144-158.

[47]Hiremath S M, Deshmukh S, Rakesh R, et al. Blind identification of radio access techniques based on time-frequency analysis and convolutional neural network[C]//TENCON 2018-2018 IEEE Region 10 Conference. IEEE, 2018: 1163-1167.

[48]Xia H, Alshathri K, Lawrence V B, et al. Cellular signal identification using convolutional neural networks: AWGN and Rayleigh fading channels[C]//2019 IEEE International Symposium on Dynamic Spectrum Access Networks (DySPAN). IEEE, 2019: 1-5.

[49]Alhazmi M H, Alymani M, Alhazmi H, et al. 5G signal identification using deep learning[C]//2020 29th Wireless and Optical Communications Conference (WOCC), 2020: 1-15.

[50]Ngo T, Kelley B, Rad P. Deep learning based prediction of signal-to-noise ratio (SNR) for LTE and 5G systems[C]//2020 8th International Conference on Wireless Networks and Mobile Communications (WINCOM). IEEE, 2020: 1-6.

[51]Wu H, Jin B, Xu Z, et al. Waveform design and signal processing for integrated radar-communication system based on frequency diversity array[J]. Digital Signal Processing, 2023: 133.

[52]Maggiolo L, Marcos D, Moser G, et al. A semisupervised CRF model for CNN-based semantic segmentation with sparse ground truth[J]. IEEE Transactions on Geoscience and Remote Sensing, 2021, 3(99): 1-15.

[53]Sunnetci K M, Kaba E, et al. Comparative parotid gland segmentation by using ResNet‐18 and MobileNetV2 based DeepLab v3+ architectures from magnetic resonance images[J]. Concurrency and Computation: Practice and Experience, 2023, 35(1): e7405.

[54]Hou C, Fu D, Hua L, et al. The recognition of multi-components signals based on semantic segmentation[J]. Wireless Networks, 2023: 1-11

[55]Wang Z, Wang J, Yang K, et al. Semantic segmentation of high-resolution remote sensing images based on a class feature attention mechanism fused with Deeplab v3+[J]. Computers & Geosciences, 2022, 158: 104969.

[56]Wang Q, Liu H, Wen T, et al. Research on Semantic Segmentation Algorithm for Multiscale Feature Images Based on Improved DeepLab v3+[C]//Journal of Physics: Conference Series. IOP Publishing, 2024, 2747(1): 012038.

[57]Gao X, Chen C, Gan Y, et al. Comparison of Extraction accuracy of Sugarcane from different resolution satellite images using Deep lab V3+ Mode[J]. The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, 2024, 48: 179-184.