正交时频空间（Orthogonal Time-Frequency-Space，OTFS）是一种新的二维调制技术，可在时频选择性信道上提供可靠的通信。在水声信道中，时间选择性衰落和频率选择性衰落都非常严重，接收机必须准确恢复因干扰而失真的OTFS信号。近年来，深度学习在通信系统中实现了比传统方法更好的性能，为设计OTFS接收端的优化方法提供了新的思路。本文使用深度学习方法对OTFS接收端的信道估计和信号检测进行优化。论文主要研究成果包括：

（1）提出了一种基于模型驱动的深度学习水声 OTFS信道估计方法。通过将基于阈值的信道估计方法与去噪卷积神经网络（Denoising Convolutional Neural Network，DnCNN）级联进行信道估计，级联中的基于阈值的算法生成初步的信道估计结果，再利用级联中的DnCNN对粗略估计结果进行进一步的去噪处理，得到准确的OTFS信道估计。在仿真和实测水声信道下进行验证，结果表明，所提方法的性能明显优于传统的基于阈值的算法以及基于正交匹配追踪的算法。

（2）提出了两种基于数据驱动的深度学习水声 OTFS信号检测方法。首先提出了基于全连接深度神经网络（Fully Connected Deep Neural Network，FC-DNN）的水声 OTFS信号检测方法，FC-DNN可以通过参数的迭代优化来拟合输入输出关系。由于FC-DNN存在非凸优化和梯度消失问题，限制了其鲁棒性，在此基础上提出了基于联合卷积和循环神经网络的水声OTFS信号检测方法，将带有跳跃连接（Skip Connection, SC）的卷积神经网络（Convolutional Neural Network，CNN）和双向长短期记忆网络（Bi-directional Long Short-Term Memory, BiLSTM）级联，以进行信号恢复，称为SC-CNN-BiLSTM。在仿真和实测水声信道下进行验证，仿真结果表明，与2D-CNN、FC-DNN和传统检测器相比，基于SC-CNN-BiLSTM的OTFS检测方法具有更低的误码率。

    Orthogonal Time-Frequency-Space (OTFS) is a new two-dimensional modulation technique that provides reliable communication on time-frequency selective channels. In the underwater acoustic channel, both time-selective fading and frequency-selective fading are severe. The receiver must recover the OTFS signal that has been distorted by interference. In recent years, deep learning can show better performance than traditional methods in communication systems, providing new ideas for designing optimization methods for OTFS receivers. In this thesis, deep learning methods are used to optimize channel estimation and signal detection for OTFS receiver. The main research achievements of the paper include:

(1) A model-driven deep learning underwater acoustic OTFS channel estimation method is proposed. By cascading the threshold-based channel estimation method with the Denoising Convolutional Neural Network (DnCNN), the threshold-based algorithm in the cascade generates preliminary channel estimation results, and then the DnCNN in the cascade is used to further denoise the rough estimation results to obtain accurate OTFS channel estimation. Verified under simulated and measured underwater acoustic channels, the results show that the performance of the proposed method is significantly better than the traditional threshold-based algorithm and the Orthogonal Matching Pursuit algorithm.

(2) Two data-driven deep learning underwater acoustic OTFS signal detection methods are proposed. Firstly, a underwater acoustic OTFS signal detection method based on Fully Connected Deep Neural Network (FC-DNN) is proposed, and FC-DNN can fit the input-output relationship through iterative optimization of parameters. Since FC-DNN has the problems of non-convex optimization and gradient vanishing, which limits its robustness, on this basis, a submarine-acoustic OTFS signal detection method based on joint convolutional and recurrent neural network is proposed, and the convolutional neural network (CNN) with skip connection (SC) and bidirectional long short-term memory network (BiLSTM) are proposed for signal recovery, called SC-CNN-BiLSTM. The simulation results show that the SC-CNN-BILSTM based OTFS detection method has a lower bit error rate than 2D-CNN, FC-DNN and traditional detectors.

[1]许肖梅. 水声通信与水声网络的发展与应用[J]. 声学技术, 2009, 28(06):811-816.

[2]王毅凡, 周密, 宋志慧. 水下无线通信技术发展研究[J]．通信技术, 2014, 6:589-594.

[3]Stojanovic M. Low Complexity OFDM Detector for Underwater Acoustic Channels[C] //OCEANS 2006, Boston, MA, USA, 2006: 1-6.

[4]张舒然, 武岩波, 朱敏, 等. 基于模拟遗传算法的水声时变多普勒频偏估计[J]. 网络新媒体技术, 2020, 9(06): 45-50.

[5]Hadani R, Rakib S, Molisch A, et al. Orthogonal time frequency space (OTFS) modulation for millimeter wave communications systems[C] //IEEE MTTS Intl Microwave Symp, 2017: 681-683.

[6]Farhang A，Rezadeheyhani A，Doylele L E，et al．Low complexity modem structure for OFDM-based orthogonal time frequency space modulation[J]. IEEE Wireless Communications Letters, 2017, 7(3): 344-347.

[7]Monk A, Hadani R. OTFS: A new generation of modulation addressing the challenges of 5G [C] //IEEE WCNC, San Francisco CA, 2017:1-6.

[8]Raviteja P, Phan K T, Hong, Y, et al．Interference Cancellation and Iterative Detection for Orthogonal Time Frequency Space Modulation[J]. IEEE Transactions on Wireless Communications, 2018, 17: 6501-6515.

[9]Raviteja P, Hong Y, Viterbo E, et al．Practical Pulse-Shaping Waveforms for Reduced-Cyclic-Prefix OTFS[J]. IEEE Transactions on Vehicular Technology, 2019, 68(1): 957-961.

[10]Rangamgari V, Tiwari S, Das S S, et al. OTFS: Interleaved OFDM with Block CP[C] //2020 National Conference on Communications (NCC), 2020: 1-6.

[11]Ding Z, Schober R, Fan P, et al. OTFS-NOMA: An Efficient Approach for Exploiting Heterogenous User Mobility Profiles[J]. IEEE Transactions on Communications, 2019, 67(11): 7950-7965.

[12]Raviteja P, Phan K T, and Hong Y. Embedded pilot-aided channel estimation for OTFS in delay-Doppler channels[J]. IEEE Transactions on Vehicular Technology, 2019, 68(5): 4906-4917.

[13]杜坤鹏, 陈文. OTFS系统中基于最小多普勒间干扰的导频设计[J]. 信息技术, 2020, 44, 342(05):18-22.

[14]Murali K R, Chockalingam A. On OTFS Modulation for High-Doppler Fading Channels[C] //2018 Information Theory and Applications Workshop (ITA), San Diego, CA, USA, 2018:1-10.

[15]Raviteja P, Viterbo E, Hong Y. OTFS Performance on Static Multipath Channels[J]. IEEE Wireless Communications Letters. 2019, 8:745-748.

[16]Tiwari S, Das S S, Rangamgari V. Low complexity LMMSE Receiver for OTFS[J]. IEEE Communications Letters. 2019, 23: 2205-2209.

[17]Raviteja P, Phan K T, Hong Y, et al. Interference Cancellation and Iterative Detection for Orthogonal Time Frequency Space Modulation[J]. IEEE transactions on communications, 2018, 17: 6501-6515.

[18]Li L, Liang Y, Fan P, et al. Low Complexity Detection Algorithms for OTFS under Rapidly Time-Varying Channel[C] //2019 IEEE 89th Vehicular Technology Conference (VTC2019-Spring), 2019: 1-5.

[19]李伶琚. 抗多普勒频移正交时频空系统性能分析与优化[D]. 重庆: 西南交通大学, 2019.

[20]Tiwari S, Das S S, Rangamgari V. Low complexity LMMSE Receiver for OTFS [J]. IEEE Communications Letters, 2019, 23(12): 2205-2209.

[21]Feng X, Esmaie L H, Wang J, et al. Underwater Acoustic Communications Based on OTFS [C] //2020 15th IEEE International Conference on Signal Processing (ICSP), IEEE, 2020: 439-444.

[22]张阳, 张群飞, 王樱洁, 等. 一种低复杂度正交时频空间调制水声通信方法[J]. 西北工业大学学报, 2021 , 39(5):954-961.

[23]Bocus M J, Doufexi A, Agrafiotis D. Performance of OFDM-based massive MIMO OTFS systems for underwater acoustic communication[J]. IET communications, 2020, 14(4): 588-593.

[24]Jing L, Zhang N, He C, et al. OTFS underwater acoustic communications based on passive time reversal[J]. Applied Acoustics, 2022, 185: 108386.

[25]Mirza G, Kien N, Gabriel P V, et al. Big data analytics, machine learning, and artificial intelligence in next-generation wireless networks[J]. IEEE access, 2018, 6:32328-32338.

[26]Li R, Zhao Z, Zhou X et al. Intelligent 5G: When cellular networks meet artificial intelligence[J]. IEEE Wireless communications, 2017, 24(5):175-183.

[27]Han S, Chih-Lin I, Li G, et al. Big data enabled mobile network design for 5G and beyond[J]. IEEE Communications Magazine, 2017, 55(9):150-157.

[28]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, 2018, 7(1): 114-117.

[29]Zhang Y, Li J, Zakharov Y, et al. Deep learning based underwater acoustic OFDM communications[J]. Applied Acoustics, 2019, 154: 53-58.

[30]Zhang J, Cao Y, Han G, et al. Deep neural network-based underwater OFDM receiver [J]. IET communications, 2019, 13(13): 1998-2002.

[31]Zhang Y, Li C, Wang H, et al. Deep learning aided OFDM receiver for underwater acoustic communications[J]. Applied Acoustics, 2022, 187: 108515.

[32]Wang B, Xu K, Zheng S, et al. A Deep Learning-Based Intelligent Receiver for Improving the Reliability of the MIMO Wireless Communication System[J]. IEEE Transactions on Reliability, 2022, 71(2): 1104-1115.

[33]Li J, Xin T, He B, et al. IQ Symbols Processing Schemes with LSTMs in OFDM System[J]. IEEE Access, 2022, 10: 70737-70745.

[34]Wang S, Yao R, Tsiftsis T A, et al. Signal Detection in Uplink Time-Varying OFDM Systems Using RNN With Bidirectional LSTM[J]. IEEE Wireless Communications Letters, 2020, 9 (11): 1947-1951.

[35]He H, Wen C -K, Jin S, et al. Deep Learning-Based Channel Estimation for Beamspace mmWave Massive MIMO Systems[J]. IEEE Wireless Communications Letters, 2018, 7(5): 852-855.

[36]Soltani M, Pourahmadi V, Mirzaei A, et al. Deep Learning-Based Channel Estimation[J]. IEEE Communications Letters, 23(4): 652-655.

[37]Liao Y, Hua Y, Cai Y, Deep Learning Based Channel Estimation Algorithm for Fast Time-Varying MIMO-OFDM Systems[J]. IEEE Communications Letters, 2020, 24(3): 572-576.

[38]NAIKOTI A, CHOCKALINGAM A. A Dnn-based OTFS Transceiver with Delay-Doppler Channel Training and IQI Compensation[C] //2021 IEEE 32nd Annual International Symposium on Personal, Indoor and Mobile Radio Communications (PIMRC). IEEE, 2021: 628-634.

[39]Li Q, Gong Y, Wang J, et al. Exploring the Performance of Receiver Algorithm in OTFS Based on CNN[C] //2022 IEEE International Conference on Communications Workshops (ICC Workshops). IEEE, 2022: 957-962.

[40]Enku Y K, Bai B, Wan F, et al. Two-Dimensional Convolutional Neural Network-Based Signal Detection for OTFS Systems[J]. IEEE Wireless Communications Letters, 2021, 10(11): 2514-2518.

[41]张静, 金石, 温朝凯. 基于人工智能的无线传输技术最新研究进展[J]. 电信科学, 2018, 4(08): 46-55.

[42]Shen W, Dai L, An J, et al. Channel estimation for orthogonal time frequency space (OTFS) massive MIMO[J], IEEE Transactions on Signal Processing, 2019, 67(16): 4204-4217.

[43]Zhang K, Zuo W, Chen Y, et al. Beyond a Gaussian Denoiser: Residual Learning of Deep CNN for Image Denoising[J]. IEEE Transactions on Image Processing, 2017, 26(7): 3142-3155.

[44]Zaheer R, Shaziya H. A Study of the Optimization Algorithms in Deep Learning[C] //2019 Third International Conference on Inventive Systems and Control (ICISC), Coimbatore, India, 2019: 536-539.

[45]Van Walree P A, Socheleau F X, Otnes R, et al. The Watermark Benchmark for Underwater Acoustic Modulation Schemes[J]. IEEE Journal of Oceanic Engineering. 2017, 42: 1007-1018.

[46]Raviteja P, Hong Y, Viterbo E, et al. Practical pulse-shaping waveforms for reduced-cyclic-prefix OTFS[J]. IEEE Transactions on Vehicular Technology. 2018, 68(1):957-961.

[47]Vaze C S, Varanasi M K. The degrees of freedom region and interference alignment for the MIMO interference channel with delayed CSI[J]. IEEE Transactions on Information Theory. 2012, 58: 4396-4417.

[48]He K, Sun J. Convolutional neural networks at constrained time cost[C] //In Proceedings of the 2015 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), Boston, MA, USA, 2015, 7: 5353-5360.