矿井信息化、智能化的发展与无线通信技术密切相关。在矿井中存在多种不同的无线通信系统并且每种通信系统采用的信号调制方式各不相同，因此要形成安全可靠的矿井通信网络系统，就必须实现不同通信系统间的信号调制方式识别。本文主要针对矿井无线信道环境研究了基于信号特征提取的调制识别方法和基于深度学习的调制识别方法，对矿井衰落信道下的BPSK、QPSK、8PSK、16PSK、16QAM、64QAM、256QAM、4PAM和GMSK九种常用的调制信号进行识别分类。

       基于信号特征提取的调制识别算法，本文通过分析信号高阶累积量与矿井小尺度Nakagami衰落的关系，计算出信号经过衰落信道的高阶累积量值。进而选取信号的四阶、六阶和八阶累积量作为特征参量并构造特征值向量。分别设计了决策树、支持向量机（SVM）和全连接神经网络作为目标分类器，实现了对矿井衰落信道下的九种调制信号识别。仿真结果表明，基于决策树模型的识别效果比SVM和全连接神经网络差，SVM和全连接神经网络的识别效果比较接近。在低信噪比下，三个模型的识别率都比较低，并且不易区分64QAM和256QAM信号。

       针对基于信号特征识别方法存在的缺陷，本文进一步提出基于深度学习的调制识别方法。分别设计了基于卷积神经网络(CNN)、长短期记忆(LSTM)网络和CLDNN (Convolutional, Long short-term memory, Fully connected Deep Neural Networks)模型，并将这些网络模型应用于矿井信道环境下的调制识别领域，通过分别研究CNN和LSTM模型的层数对识别率的影响，进而设计出了性能较好的CLDNN模型。仿真结果表明，相比较单一的网络模型，基于组合神经网络模型的识别率更高，与基于信号特征的调制识别方法相比，提高了低信噪比下的识别率。

      本文进一步提出了基于注意力机制改进的AM-CLDNN网络模型，通过不同规模的训练数据集进行对深度学习网络模型性能测试。其结果表明，基于注意力机制改进的AM-CLDNN模型效果最优，并且有效的提高了低信噪比下的识别性能。这说明注意力机制模型和深度学习算法的结合使用，能有效提高复杂信道下的调制信号识别性能。

The development of mine informatization and intelligence is closely related to wireless communication technology. There are many different wireless communication systems in the mine, and the signal modulation methods used by each communication system are different. Therefore, to form a safe and reliable mine communication network system, it is necessary to realize the recognition of signal modulation modes between different communication systems. This dissertation studies the recognition method based on artificial extraction of signal features and the modulation recognition method based on deep learning for the mine wireless channel environment. This thesis identifies and classifies the BPSK, QPSK, 8PSK, 16PSK, 16QAM, 64QAM, 256QAM, 4PAM and GMSK modulation signals in mine fading channel.

Modulation recognition algorithm based on signal feature extraction. In this thesis, by analyzing the relationship between the high-order cumulant of the signal and the small-scale Nakagami fading of the mine, the calculation expression of the high-order cumulant of the signal passing through the fading channel is derived. Then select the fourth, sixth and eighth-order cumulants as the characteristic parameters and construct the eigenvalue vector. A decision tree, a support vector machine (SVM) and a fully connected neural network are designed as target classifiers to realize the classification and recognition of nine kinds of modulation signals in the fading channel of the mine. The simulation results show that the recognition effect of decision tree classifier is worse than SVM and fully connected neural network classifier. The recognition performance of the classifier based on SVM and fully connected neural network is equivalent, but the recognition effect is not ideal under low signal-to-noise ratio, and it is difficult to distinguish 64QAM and 256QAM signals.

Aiming at the shortcomings of traditional modulation recognition methods, this thesis further proposes a modulation recognition method based on deep learning. The models based on Convolutional Neural Networks, Long Short-term Memory Networks and CLDNN (Convolutional, Long Short-term Memory, Fully Connected Deep Neural Networks) are designed respectively. These network models are applied to the field of modulation recognition in the mine channel environment, and the CLDNN model with better performance is designed by separately studying the influence of the number of layers of CNN and LSTM models on the recognition rate. The simulation results show that compared with a single network model, the recognition rate based on the combined neural network model is higher. Compared with the modulation recognition method based on signal characteristics, the recognition rate under low signal-to-noise ratio is improved.

This thesis further proposes an improved AM-CLDNN network model based on the attention mechanism, and tests the performance of the deep learning network model through training data sets of different sizes. The results show that the improved AM-CLDNN model based on the attention mechanism has the best effect. And it effectively improves the recognition performance under low signal-to-noise ratio. This shows that the combined use of the attention mechanism model and the deep learning algorithm can effectively improve the modulation signal recognition performance under complex channels.

[1] 王国法,王虹,任怀伟等.智慧煤矿2025情景目标和发展路径[J].煤炭学报, 2018,43(02): 295-305.

[2] 范京道,李川,闫振国.融合5G技术生态的智能煤矿总体架构及核心场景[J].煤炭学报,2020,45(06):1949-1958.

[3] 孙继平.智慧矿山与5G和WiFi6 [J]. 工矿自动化,2019,45(10):1-4.

[4] 李小四,马建民,王莹莹.智慧矿山建设的演进及发展趋势[J].工矿自动化,2019,45(09):65-69.

[5] Li B, Li S, Hou J, et al. A Bayesian approach for adaptively modulated signals recognition in next-generation communications[J]. IEEE Transactions on Signal Processing, 2015, 63(16): 4359-4372.

[6] Lyu Z, Wang Y, Li W, et al. Robust automatic modulation classification based on convolutional and recurrent fusion network[J]. Physical Communication, 2020: 101213.

[7] 张思成,林云,涂涯,Shiwen Mao.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.

[8] Ding G, Wang J, Wu Q, et al. Robust spectrum sensing with crowd sensors[J]. IEEE Transactions on Communications, 2014, 62(9): 3129-3143.

[9] 龚晓洁. 基于支持向量机的调制方式识别算法的研究[D].南京邮电大学,2011.

[10] 王艳芬,王亮,孙彦景,张亮,徐华,潘东跃.面向应急救援的多输入单输出磁感应透地通信[J].工矿自动化,2019,45(06):10-15.

[11] 高则海. 矿井巷道无线信道特性分析与建模仿真[D].安徽理工大学,2016.

[12] 李蓉. 基于SVM矿井环境无线多模信号的检测识别[D].西安科技大学,2019.

[13] Nandi A K, Azzouz E. Algorithms for automatic modulation recognition of communication signals[J]. IEEE Transactions on communications, 1998, 46(4): 431-436.

[14] Chen W, Xie Z, Ma L, et al. A faster maximum-likelihood modulation classification in flat fading non-Gaussian channels[J]. IEEE Communications Letters, 2019, 23(3): 454-457.

[15] Ozdemir O, Li R, Varshney P K. Hybrid maximum likelihood modulation classification using multiple radios[J]. IEEE Communications Letters, 2013, 17(10): 1889-1892.

[16] Yuan Y, Zhao P, Wang B, et al. Hybrid maximum likelihood modulation classification for continuous phase modulations[J]. IEEE Communications Letters, 2016, 20(3): 450-453.

[17] Ghasemzadeh P, Banerjee S, Hempel M, et al. A novel deep learning and polar transformation framework for an adaptive automatic modulation classification[J]. IEEE Transactions on Vehicular Technology, 2020, 69(11): 13243-13258.

[18] 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: 257-261.

[19] 龚安民,王炳和,曲毅.基于同步压缩小波变换的通信信号调制识别[J].电光与控制,2015,22(12):50-53+71.

[20] Zhang J, Wang F, Dobre O A, et al. Specific emitter identification via Hilbert–Huang transform in single-hop and relaying scenarios[J]. IEEE Transactions on Information Forensics and Security, 2016, 11(6): 1192-1205.

[21] Zhao Y, Zou Z, Wu L, et al. Frequency Detection Algorithm for Frequency Diversity Signal Based on STFT[C]//2015 Fifth International Conference on Instrumentation and Measurement, Computer, Communication and Control (IMCCC). IEEE, 2015: 790-793.

[22] Kumar Y, Sheoran M, Jajoo G, et al. Automatic modulation classification based on constellation density using deep learning[J]. IEEE Communications Letters, 2020, 24(6): 1275-1278.

[23] 张华娣,楼华勋.MQAM信号调制方式自动识别方法[J].通信学报,2019,40(08):200-211.

[24] 李东波,欧阳喜.一种抗频偏的常见幅相调制信号识别算法[J].信息工程大学学报,2018,19(01):42-46.

[25] 谭晓衡,褚国星,张雪静,杨扬.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(01):171-177.

[26] 张利,李青.基于高阶累积量的调制识别算法的研究[J].信息工程大学学报,2017,18(04):403-408.

[27] Yoo D S, Lim J, Kang M H. ATSC digital television signal detection with spectral correlation density[J]. Journal of Communications and Networks, 2014, 16(6): 600-612.

[28] 赵雄文,郭春霞,李景春.基于高阶累积量和循环谱的信号调制方式混合识别算法[J].电子与信息学报,2016,38(03):674-680.

[29] Schmidhuber J. Deep learning in neural networks: An overview[J]. Neural networks, 2015, 61: 85-117.

[30] Krizhevsky A, Sutskever I, Hinton G E. Imagenet classification with deep convolutional neural networks[C]//Advances in neural information processing systems. 2012: 1097-1105.

[31] Sainath T N, Weiss R J, Senior A, et al. Learning the speech front-end with raw waveform CLDNNs[C]//Sixteenth Annual Conference of the International Speech Communication Association. 2015.

[32] O’Shea T, Hoydis J. An introduction to deep learning for the physical layer[J]. IEEE Transactions on Cognitive Communications and Networking, 2017, 3(4): 563-575.

[33] Liu M, Song T, Hu J, et al. Deep learning-inspired message passing algorithm for efficient resource allocation in cognitive radio networks[J]. IEEE Transactions on Vehicular Technology, 2018, 68(1): 641-653.

[34] Zhu X, Fujii T. Modulation classification for cognitive radios using stacked denoising autoencoders[J]. International Journal of Satellite Communications and Networking, 2017, 35(5): 517-531.

[35] Zhang Z, Hua Z, Liu Y. Modulation classification in multipath fading channels using sixth-order cumulants and stacked convolutional auto-encoders[J]. IET communications, 2017, 11(6): 910-915.

[36] 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, 30(3): 718-727.

[37] Zeng Y, Zhang M, Han F, et al. Spectrum analysis and convolutional neural network for automatic modulation recognition[J]. IEEE Wireless Communications Letters, 2019, 8(3): 929-932.

[38] Li R, Li L, Yang S, et al. Robust automated VHF modulation recognition based on deep convolutional neural networks[J]. IEEE Communications Letters, 2018, 22(5): 946-949.

[39] 刘明骞,郑诗斐,李兵兵.基于深度学习的MPSK信号调制识别[J].国防科技大学学报,2019,41(05):153-158.

[40] O'shea T J, West N. Radio machine learning dataset generation with gnu radio[C]//Proceedings of the GNU Radio Conference. 2016, 1(1).

[41] O’Shea T J, Corgan J, Clancy T C. Convolutional radio modulation recognition networks[C]//International conference on engineering applications of neural networks. Springer, Cham, 2016: 213-226.

[42] 郭有为. 基于深度学习的自动调制识别技术研究[D].中国工程物理研究院,2018.

[43] 吴傲.基于大数据的信号调制方式识别研究及软件设计[D].北京邮电大学,2019.

[44] Wang Y, Liu M, Yang J, et al. Data-driven deep learning for automatic modulation recognition in cognitive radios[J]. IEEE Transactions on Vehicular Technology, 2019, 68(4): 4074-4077.

[45] Ranjan A, Misra P, Sahu H B. Experimental measurements and channel modeling for wireless communication networks in underground mine environments[C]//2017 11th European Conference on Antennas and Propagation (EUCAP). IEEE, 2017: 1345-1349.

[46] 杨维, 李滢, 孙继平. 类矩形矿井巷道中UHF宽带电磁波统计信道建模[J]. 煤炭学报, 2008, (04): 467-472.

[47] 王艳芬. 矿井超宽带无线通信信道模型研究[D].中国矿业大学,2009.

[48] 王艳芬, 于洪珍, 张传祥. 矿井超宽带复合衰落信道建模及仿真[J]. 电波科学学报, 2010, 25(04): 805-812.

[49] 张剑,黄媛媛,周兴建,卢建川.GMSK信号的后解码相干解调[J].电讯技术,2017,57(04):393-396.

[50] 史蕴豪,许华,刘英辉.一种基于伪标签半监督学习的小样本调制识别算法[J].西北工业大学学报,2020,38(05):1074-1083.

[51] Ali A, Yangyu F. Unsupervised feature learning and automatic modulation classification using deep learning model[J]. Physical Communication, 2017, 25: 75-84.

[52] 陈蕙心. 通信信号调制识别技术研究[D].西安电子科技大学,2017.

[53] 李正权,林媛,李梦雅,刘洋,吴琼,邢松.基于判别式受限玻尔兹曼机的数字调制识别[J].通信学报,2021,42(02):81-91.

[54] 雷志坤.基于深度学习的调制识别技术研究[D].四川:电子科技大学,2019.

[55] Wang Y, Gui J, Yin Y, et al. Automatic modulation classification for MIMO systems via deep learning and zero-forcing equalization[J]. IEEE Transactions on Vehicular Technology, 2020, 69(5): 5688-5692.

[56] Wang Y, Yang J, Liu M, et al. LightAMC: Lightweight automatic modulation classification via deep learning and compressive sensing[J]. IEEE Transactions on Vehicular Technology, 2020, 69(3): 3491-3495.

[57] Sak H, Senior A W, Beaufays F. Long short-term memory recurrent neural network architectures for large scale acoustic modeling[J]. Computer Science, 2014: 338-342.

[58] Sainath T N, Vinyals O, Senior A, et al. Convolutional, long short-term memory, fully connected deep neural networks[C]//2015 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP). IEEE, 2015: 4580-4584.

[59] K. Xu, J. Ba, R. Kiros, K. Cho, A. Courville, R. Salakhudinov, R. Zemel, and Y. Bengio, ‘‘Show, attend and tell: Neural image caption generation with visual attention,’’ in Proc. Int. Conf. Mach. Learn., 2015, pp. 2048–2057.

[60] Denil M., Bazzani L., Larochelle H., et al. Learning where to attend with deep architectures for image tracking[J]. Neural Computation, 2012, 24(8): 2151-2184.

[61] Yang Z, Yang D, Dyer C, et al. Hierarchical attention networks for document classification[C]//Proceedings of the 2016 conference of the North American chapter of the association for computational linguistics: human language technologies. 2016: 1480-1489.

[62] Su Jinsong, Zeng Jiali, Xiong Deyi, et al. A hierarchy-to-sequence attentional neural machine translation model[J]. IEEE/ACM Transactions on Audio, Speech, and Language Processing, 2018, 26(3): 623-632.