煤矿井下环境复杂，信号在传输过程中受到噪声以及其它信道衰落干扰会导致严重失真，无线通信中 OFDM 系统接收机的性能极大程度上取决于信号检测；当前传统信号检测方法在复杂度和误比特率性能之间难以取得平衡。本文提出利用深度学习技术研究基于矿井环境下 OFDM 接收端信号检测算法，具体研究内容如下：

       针对在 OFDM 系统中由于信道的多径衰落和噪声带来的接收端出现信号失真、频偏等问题，本文利用 DenseNet 网络和 LSTM 网络各自的优势设计了一种基于 DenseNet LSTM 的信号检测网络模型，在 DenseNet 网络最后一个最大池化后以级联的方式加入LSTM 网络。DenseNet 网络用于提取信号的空间和频率特征信息以及有效利用网络中的参数；LSTM网络用来处理序列数据和捕捉OFDM信号数据的时序特征。并分别在不同信道条件、不同网络模型、不同编码和调制方式以及不同的比特流位数情况下对该信号检测算法模型进行仿真验证分析，结果表明，基于 DenseNet-LSTM 的神经网络检测算法模型相比于传统 OFDM 信号检测算法以及单个网络 LSTM、DenseNet 算法模型可以有效降低矿井 OFDM 通信的误比特率。

       现有的高精度 DenseNet-LSTM 网络模型在设计过程中没有考虑模型的大小和复杂 性，无法以一个高效的网络模型用于未来计算和部署资源有限的物联网设备，针对该问题设计了一种基于 CNN-GRU 信号检测网络模型，使用低层数的基本 CNN 网络和 GRU网络级联并削减该模型的尺度，并分别对 DenseNet-LSTM 网络中含有 4 个 Dense Block块的 DenseNet 和 LSTM 神经网络进行结构优化。分别在不同的网络模型、不同的调制方式和不同的比特流位数以及有无导频和循环前缀（CP）等情况下对该信号检测算法模型进行仿真验证，结果表明，相较于 DenseNet-LSTM 的神经网络检测算法模型，在牺牲一定的精度条件下，实现了训练速度的提升以及网络结构和参数的优化。同时还验证了基于 CNN-GRU 神经网络的检测算法仍然比单个网络 DenseNet、GRU 以及传统信号检测算法优势明显，有效降低了误比特率。

       The underground environment of coal mine is complex, and the signal will be seriously distorted by noise and other channel fading interference during transmission. The performance of OFDM system receiver in wireless communication depends largely on signal detection. The current traditional signal detection methods are difficult to achieve a balance between complexity and bit error rate performance. This paper proposes to use deep learning technology to study OFDM receiver signal detection algorithm based on mine environment. The specific research contents are as follows:

        Aiming at the problems of signal distortion and frequency offset at the receiving end caused by multipath fading and noise in OFDM system, this thesis designs a signal detection network model based on DenseNet-LSTM by using the advantages of DenseNet network and LSTM network. After the last maximum pooling of DenseNet network, LSTM network is added in cascade. The DenseNet network is used to extract the spatial and frequency feature information of the signal and effectively use the parameters in the network; the LSTM network is used to process sequence data and capture the timing characteristics of OFDM signal data. The signal detection algorithm model is simulated and verified under different channel conditions, different network models, different coding and modulation methods and different bit stream bits. The results show that the neural network detection algorithm model based on DenseNet-LSTM can effectively reduce the bit error rate of mine OFDM communication compared with the traditional OFDM signal detection algorithm and single network LSTM and DenseNet algorithm models.

         The existing high-precision DenseNet-LSTM network model does not consider the size and complexity of the model in the design process, and it is impossible to use an efficient network model for future computing and deployment of IoT devices with limited resources. Aiming at this problem, a signal detection network model based on CNN-GRU is designed. The basic CNN network and GRU network with low layers are cascaded and the scale of the model is reduced. The DenseNet and LSTM neural networks with four Dense Blocks in the DenseNet-LSTM Network are optimized. The signal detection algorithm model is simulated and verified under different network models, different modulation modes, different bit stream bits, with or without pilot and cyclic prefix (CP). The results show that compared with the neural network detection algorithm model of DenseNet-LSTM, the training speed is improved and the network structure and parameters are optimized at the expense of a certain accuracy. It is also verified that the detection algorithm based on CNN-GRU neural network still has obvious advantages over single network DenseNet, GRU and traditional signal detection algorithms, which effectively reduces the bit error rate.

[1] 孙继平. 煤矿智能化与矿用 5G[J].工矿自动化,2020,46(08):1-7．

[2] 王国法,赵国瑞,任怀伟. 智慧煤矿与智能化开采关键核心技术分析[ J] . 煤炭学

报,2019,44 (1) :34 -41.

[3] 彭苏萍. 我国煤矿安全高效开采地质保障系统研究现状及展望[J]. 煤炭学报,

2020,45(7):2331-2345.

[4] 霍振龙, 张袁浩. 5G 通信技术及其在煤矿的应用构想[J]. 工矿自动化, 2020, 46(3): 1-5.

[5] 霍振龙. 矿井无线通信系统现状与发展趋势[J]. 工矿自动化，2022，48(6): 1-5.

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

[7] LeCun Y, Bengio Y, Hinton G. Deep learning[J]. nature, 2015, 521(7553): 436-444.

[8] Zheng S , Chen S , Yang X . DeepReceiver: A Deep Learning-Based Intelligent Receiver for Wireless Communications in the Physical Layer[J]. IEEE Transactions on Cognitive Communications and Networking,2020,PP(99):1-1.

[9] 潘磊, 陈岚, 朱胜利, 等. 一种正交频分复用定时同步算法及其硬件实现优化[J]. 吉林大学学报 (工学版), 2022, 52(11): 2728-2734.

[10] 魏景新,冉小英.矿井 OFDM 无线通信系统信道仿真[J].辽宁工程技术大学学报(自然科学版),2015,34(12):1345-1349.

[11]郑建宏, 李玉菱, 龚明苇, 等. 基于信噪比排序的 MIMO-OFDM 信号检测方法[J]. 重庆邮电大学学报: 自然科学版, 2017, 29(4): 427-432.

[12]何江. OFDM 系统的信道估计研究与实现[D]. 西南交通大学, 2017.

[13]谢斌,杨丽清,陈琴.基于 EMD-SVD 差分谱的 DWT 域 LMMSE 自适应信道估计算法

[J].计算机应用,2016,36(11):3033-3038.

[14]Li H, Wang Z, Wang H. An energy-efficient power allocation scheme for Massive MIMO systems with imperfect CSI[J]. Digital Signal Processing, 2021, 112: 102964.

[15]赵太飞,刘龙飞,王晶,杨黎洋.无线紫外光散射通信中的改进 CMA-FSE 盲均衡算法[J]. 通信学报,2019,40(03):102-108.

[16]时伟. 短 CP 场景中 OFDM 解调技术研究[D]. 东南大学, 2019.

[17]Al-Naffouri T Y, Quadeer A A. Cyclic prefix based enhanced data recovery in OFDM[J]. IEEE Transactions on Signal Processing, 2010, 58(6): 3406-3410.

[18]Rathinakumar S, Radunovic B, Marina M K. CPRecycle: Recycling cyclic prefix for versatile interference mitigation in OFDM based wireless systems[C]//Proceedings of the 12th International on Conference on emerging Networking EXperiments and Technologies 2016: 67-81.

[19]Yang J, Guo Q, Huang D D, et al. A factor graph approach to exploiting cyclic prefix for equalization in OFDM systems[J]. IEEE transactions on communications, 2013, 61(12): 4972-4983.

[20]Xu L, Yang J, Huang D, et al. Exploiting cyclic prefix for Turbo-OFDM receiver design[J]. IEEE Access, 2017, 5: 15762-15775.

[21]刘留,张建华,樊圆圆,于力,张嘉驰.机器学习在信道建模中的应用综述[J/OL].通信学报:1-20.

[22]Zheng B, You C, Zhang R. Fast Channel Estimation for IRS-Assisted OFDM:

10.1109/LWC.2020.3038434[P]. 2020.

[23]Kim M, Lee W, Cho D H. A novel PAPR reduction scheme for OFDM system based on deep learning[J]. IEEE Communications Letters, 2017, 22(3): 510-513.

[24]Ali A, Chen B, Raza W, et al. Reduction of PAPR by Convolutional Neural Network with Soft Feed-back in an Underwater Acoustic OFDM Communication[C]// 2021 International Bhurban Conference on Applied Sciences and Technologies (IBCAST). IEEE, 2021: 899-904.

[25]Ye H, Li G Y, Juang B H. Power of deep learning for channel estimation and signal detection in OFDM systems[J]. IEEE Wireless Communications Letters, 2017, 7(1): 114-117.

[26]Huang G, Liu Z, Van Der Maaten L, et al. Densely connected convolutional

networks[C]//Proceedings of the IEEE conference on computer vision and pattern recognition. 2017: 4700-4708.

[27]Balevi E, Andrews J G. One-bit OFDM receivers via deep learning[J]. IEEE Transactions on Communications, 2019, 67(6): 4326-4336.

[28]Al-Baidhani A, Fan H H. Learning for detection: A deep learning wireless communication receiver over Rayleigh fading channels[C]//2019 International Conference on Computing, Networking and Communications (ICNC). IEEE, 2019: 6-10.

[29]Gao X, Jin S, Wen C K, et al. ComNet: Combination of Deep Learning and Expert Knowledge in OFDM Receivers[J]. Communications Letters, IEEE, 2018, 22(12):2627-2630.

[30]Raj V, Kalyani S. Backpropagating through the air: Deep learning at physical layer without channel models[J]. IEEE Communications Letters, 2018, 22(11): 2278-2281.

[31]Liang F, Shen C, Wu F. An iterative BP-CNN architecture for channel decoding[J]. IEEE Journal of Selected Topics in Signal Processing, 2018, 12(1): 144-159.

[32]Trabelsi C, Bilaniuk O, Zhang Y, et al. Deep Complex Networks[J]. 2017.

[33]Guberman N. On Complex Valued Convolutional Neural Networks[J]. 2016.

[34]Krzyston J, Bhattacharjea R, Stark A. Modulation Pattern Detection Using Complex Convolutions in Deep Learning[J]. 2020.

[35]Zhao Z , Vuran M C , Guo F , et al. Deep-Waveform: A Learned OFDM Receiver Based on Deep Complex Convolutional Networks[J]. 2018.

[36]Bai Q, Wang J, Zhang Y, et al. Deep Learning-Based Channel Estimation Algorithm Over Time Selective Fading Channels[J]. IEEE Transactions on Cognitive Communications and Networking, 2020, 6(1):125-134.

[37]常代娜,周杰.基于深度学习算法 OFDM 信号检测[J].东南大学学报(自然科学

版),2020,50(05):912-917.

[38]Yi X, Zhong C. Deep learning for joint channel estimation and signal detection in OFDM systems[J]. IEEE Communications Letters, 2020, 24(12): 2780-2784.

[39]Emir A, Kara F, Kaya H, et al. Deep learning-based flexible joint channel estimation and signal detection of multi-user OFDM-NOMA[J]. Physical Communication, 2021, 48: 101443.

[40]Ratnam D V, Rao K N. Bi-LSTM based deep learning method for 5G signal detection and channel estimation[J]. AIMS Electronics and Electrical Engineering, 2021, 5(4): 334-341..

[41]Wu T, Fan S, Di C, et al. Signal Detection for GSTFIM Systems with Carrier Frequency Offset[J]. Sensors, 2022, 22(7): 2548..

[42]Zhang D, Wang S, Niu K, et al. Transformer-Based Detector for OFDM With Index Modulation[J]. IEEE Communications Letters, 2022, 26(6): 1313-1317.

[43]季策,宋博翰,耿蓉.快时变信道下基于深度学习的 OFDM系统信道估计[J].系统工程与电子技术:1-8[2023-03-27].

[44]王义元,常俊,卢中奎,余福慧,魏家齐.深度学习辅助的 5G OFDM 系统的信道估计

[J/OL].电讯技术:1-8[2023-03-27].

[45]邢隆,徐永海,李国权,林金朝.基于深度学习的 MIMO 系统信道估计算法[J].重庆邮电大学学报(自然科学版),2022,34(04):685-693.

[46]张玉, 杨维. 矿井巷道综合衰减模型与无线传输误码率性能综合比较[J].煤炭学报,

2011,36(09):1581-1586.

[47]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.

[48]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

[49]王艳芬. 矿井特殊环境下的超宽带无线通信信道模型研究[M]. 徐州: 中国矿业大学 出版社. 2012: 20.

[50]靖永志, 鲁林海, 冯伟, 等. 基于 OFDM 的无线信号与电能反向同步传输方法[J]. 电子与信息学报, 2022, 45: 1-11.

[51]严春满, 王铖. 卷积神经网络模型发展及应用 [J]. 计算机科学与探索 , 2021, 15(01):27-46.

[52]张玉宏. 深度学习之美[M].北京：电子工业出版社, 2018.

[53]刘颖, 杨鹏飞, 张立军, 等. 前馈神经网络和循环神经网络的鲁棒性验证综述[J]. 软件学报, 2022, 34(7): 0-0.

[54]聂倩, 杨丽花, 呼博, 等. 基扩展模型下基于 LSTM 神经网络的时变信道预测方法[J]. 系统工程与电子技术, 2022, 44(9): 2971-2977.