多制式水声通信接收机智能解译网络设计

刘兰军; 成子宁; 陈家林; 黎明; 刘鸿浩

doi:10.11993/j.issn.2096-3920.2024-0174

多制式水声通信接收机智能解译网络设计

doi: 10.11993/j.issn.2096-3920.2024-0174

刘兰军^{1, 2,},
成子宁^1,,
陈家林^{1, 2, ,},
黎明^{1, 2,},
刘鸿浩^1,

1.
中国海洋大学工程学院, 山东青岛, 266100
2.
海洋智能装备技术山东省工程研究中心, 山东青岛, 266100

基金项目: 国家自然科学基金重点项目资助(61431005).

详细信息

作者简介:
刘兰军(1979-), 男, 博士, 副教授, 主要研究方向为嵌入式技术与智能仪器、水下精密实时感知、智能信息处理与系统

通讯作者:
陈家林(1989-), 男, 硕士, 工程师, 主要研究方向为嵌入式技术与智能仪器、水下精密实时感知.

中图分类号: TJ630.34; U674.94
计量
- 文章访问数: 77
- HTML全文浏览量: 17
- PDF下载量: 15
- 被引次数: 0
出版历程
- 收稿日期: 2024-12-20
- 修回日期: 2025-03-06
- 录用日期: 2025-03-10
- 网络出版日期: 2025-04-09

Design of Intelligent Interpretation Network for Underwater Acoustic Communication Receiver with Multiple Signal Modulations

LIU Lanjun^{1, 2
,},
CHENG Zining^1
,,
CHEN Jialin^{1, 2
, ,},
LI Ming^{1, 2
,},
LIU Honghao^1
,

1.
College of Engineering, Ocean University of China, Qingdao 266100, China
2.
Shandong Provincial Engineering Research Center for Marine Intelligent Equipment Technology, Qingdao 266100, China

摘要

摘要: 针对复杂应用场景的信道自适应高质量水声通信需求, 提出了一种支持正交频分复用(OFDM)、单载波调制(SCM)、多载波频域扩频(MC-FDSS)和单载波时域扩频(SC-TDSS) 等4种信号调制方式的多制式水声通信接收机智能解译网络设计方案, 分别采用基于全连接深度神经网络(FC-DNN)和长短期记忆(LSTM)网络的智能解译模块代替传统的信道估计和信道均衡模块, 针对非扩频和扩频信号调制方式设计了易于并行扩展的深度学习网络结构。基于5种典型时变信道模型进行了网络训练和测试, 测试结果表明, 设计的2种智能解译网络相较于传统的最小二乘(LS)估计+迫零均衡、LS估计+最小均方误差信道估计均衡方法, 系统性能均有较大提升。在信噪比5 dB条件下, OFDM和SCM非扩频信号调制方式的系统误码率分别降低了约10倍和100倍; 信噪比−5 dB条件下, MC-FDSS和SC-TDSS扩频信号调制方式的系统误码率分别降低了约100倍和1 000倍; 设计的2种智能解译网络的系统性能相当, 均具有良好的泛化性能, 基于FC-DNN的智能解译网络的计算复杂度较低。
- 水声通信 /
- 深度学习 /
- 全连接深度神经网络 /
- 长短期记忆 /
- 多制式
Abstract: An intelligent interpretation network design scheme for underwater acoustic communication receivers with multiple signal modulations was proposed to meet the requirements of channel adaptive underwater acoustic high-quality communication in complex application scenarios. It supported four signal modulation methods including orthogonal frequency division multiplexing(OFDM), single carrier modulation(SCM), multi carrier frequency domain spread spectrum(MC-FDSS), and single carrier with time domain spread spectrum(SC-TDSS). Intelligent interpretation modules based on fully-connected deep neural network(FC-DNN) and long short-term memory(LSTM) networks were used to replace traditional channel estimation and channel equalization modules. A deep learning network structure that facilitated parallel expansion was designed for non-spread spectrum and spread spectrum signal modulation methods. Network training and testing were conducted based on five typical time-varying channel models. The test results show that the two designed intelligent interpretation networks have significantly improved system performance compared to traditional least squares(LS) estimation + zero-forcing(ZF) equalization and LS estimation + minimum mean squared error(MMSE) channel estimation equalization methods. Under a signal-to-noise ratio of 5 dB, the system error rates of OFDM and SCM non-spread spectrum signal modulation methods are reduced by about 10 times and 100 times, respectively. Under a signal-to-noise ratio of −5 dB, the system error rates of MC-FDSS and SC-TDSS spread spectrum signal modulation methods are reduced by about 100 times and 1 000 times, respectively. The system performance of the two designed intelligent interpretation networks is comparable, and they both have good generalization performance. The computational complexity of the intelligent interpretation network based on FC-DNN is relatively low.
- underwater acoustic communication /
- deep learning /
- fully-connected deep neural network /
- long short-term memory /
- multiple signal modulation

HTML全文

图 1 智能多制式水声通信系统

Figure 1. Intelligent underwater acoustic communication system with multiple signal modulations

下载: 全尺寸图片幻灯片

图 2 4种信号调制方式的信号处理

Figure 2. Signal processing of four signal modulations

下载: 全尺寸图片幻灯片

图 3 数据帧结构

Figure 3. Data frame structure

下载: 全尺寸图片幻灯片

图 4 FC-DNN网络结构

Figure 4. FC-DNN network structure

下载: 全尺寸图片幻灯片

图 5 LSTM网络单元结构

Figure 5. Unit structure of LSTM network

下载: 全尺寸图片幻灯片

图 6 智能多制式接收机结构

Figure 6. Structure of intelligent receiver with multiple signal modulations

下载: 全尺寸图片幻灯片

图 7 智能解译网络结构

Figure 7. Intelligent signal interpretation network architecture

下载: 全尺寸图片幻灯片

图 8 5种信道冲激响应

Figure 8. Impulse response for the five channels

下载: 全尺寸图片幻灯片

图 9 网络训练RMSE曲线

Figure 9. RMSE curves of network training

下载: 全尺寸图片幻灯片

图 10 CH1信道下智能解译网络与传统方法对比

Figure 10. Comparison between intelligent interpretation network and traditional methods in CH1 channel

下载: 全尺寸图片幻灯片

图 11 CH1信道下智能解译网络性能测试结果

Figure 11. Test results of intelligent interpretation network in CH1 channel

下载: 全尺寸图片幻灯片

图 12 CH2与CH3信道下智能解译网络性能测试结果

Figure 12. Test results of intelligent interpretation network in CH2 and CH3 channel

下载: 全尺寸图片幻灯片

图 13 CH4信道下智能解译网络的性能测试结果

Figure 13. Test results of intelligent interpretation network in CH4 channel

下载: 全尺寸图片幻灯片

图 14 CH5信道下智能解译网络性能测试结果

Figure 14. Test results of intelligent interpretation network in CH5 channel

下载: 全尺寸图片幻灯片

表 1 水声通信系统参数

Table 1. Parameters of underwater acoustic communication system

参数名称	OFDM	SCM	MC-FDSS	SC-TDSS
数字调制	QPSK	QPSK	QPSK	QPSK
频域导频数	56	—	56	—
子载波数	64	1	64	1
扩频码长度	—	—	16	16
通信速率/(kbit/s)	39.36	4.92	2.46	0.31
带宽/kHz	21~27	21~27	21~27	21~27
采样频率/kHz	96	96	96	96

下载: 导出CSV

表 2 非扩频信号FC-DNN网络结构

Table 2. FC-DNN network structure for none spread spectrum signals

类型	激活函数	神经元个数
输入层	—	256
隐藏层1	ReLU	500
隐藏层2	ReLU	250
隐藏层3	ReLU	120
输出层	sigmoid	16

下载: 导出CSV

表 3 扩频信号FC-DNN网络结构

Table 3. FC-DNN network structure for spread spectrum signals

类型	激活函数	神经元个数
输入层	—	256
隐藏层1	ReLU	500
隐藏层2	ReLU	250
输出层	sigmoid	8

下载: 导出CSV

表 4 LSTM网络结构

Table 4. LSTM network architecture

类型	激活函数	神经元个数
输入层	—	256
LSTM层1	tanh / sigmoid	128
LSTM层2	tanh / sigmoid	64
输出层	sigmoid	16(非扩频)/ 8(扩频)

下载: 导出CSV

表 5 水声信道模型参数

Table 5. Underwater acoustic channel model parameters

信道	H/m	D/m	位置	最大时延/ms
CH1	100	1 000	Ⅰ	30
CH2	100	1 000	Ⅱ	28
CH3	100	1 000	Ⅲ	28
CH4	100	5 000	Ⅱ	48
CH5	500	1 000	Ⅱ	21

下载: 导出CSV

表 6 神经网络训练参数

Table 6. Training parameters for neural networks

参数	设置
训练信噪比/dB	25
测试信噪比/dB	−30~10
优化算法	Adam
训练数据集	CH1信道每种调制35 000帧
验证数据集	CH1信道每种调制15 000帧
测试数据集	5种信道每种信噪比2 000帧
验证集评估频率	500
训练最大轮数	50(非扩频)/20(扩频)
每轮迭代次数	700
学习率	初始0.001, 两轮后0.000 8
L2正则化	0.002

下载: 导出CSV

表 7 软硬件环境

Table 7. Software and hardware environment

配置项目	参数
CPU	12^th Gen Intel(R) Core(TM) i7-12700
内存	32 GB
GPU	NVIDIA Geforce RTX 4060Ti
显存	24 GB
仿真软件	MATLAB 2022b

下载: 导出CSV

表 8 网络复杂度

Table 8. Complexity of neural networks

神经网络类型	参数量	存储/MB	FLOPS
FC-DNN非扩频	2.29×10⁶	8.52	9.12×10⁶
FC-DNN扩频	0.26×10⁶	0.97	1.02×10⁶
LSTM非扩频	1.98×10⁶	7.46	2.20×10⁷
LSTM扩频	0.25×10⁶	0.93	3.64×10⁶

下载: 导出CSV

参考文献(18)

[1]	朱敏, 武岩波. 水声通信技术进展[J]. 中国科学院院刊, 2019, 34(3): 289-296. ZHU M, WU Y B. Development of underwater acoustic communication technology[J]. Bulletin of Chinese Academy of Sciences, 2019, 34(3): 289-296.
[2]	LIU L, SUN D, ZHANG Y. A family of sparse group lasso RLS algorithms with adaptive regularization parameters for adaptive decision feedback equalizer in the underwater acoustic communication system[J]. Physical Communication, 2017, 23: 114-124. doi: 10.1016/j.phycom.2017.03.005
[3]	LI B, YANG H, LIU G, et al. A new joint channel equalization and estimation algorithm for underwater acoustic channels[J]. Eurasip Journal on Wireless Communications and Networking, 2017, 169: 1-6.
[4]	PRANITHA B, ANJANEYULU L. Analysis of underwater acoustic communication system using equalization technique for ISI reduction[J]. Procedia Computer Science, 2020, 167: 1128-1138. doi: 10.1016/j.procs.2020.03.415
[5]	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.
[6]	朱雨男, 王彪, 张岑. 基于深度神经网络的水声FBMC通信信号检测方法[J]. 声学技术, 2021, 40(2): 199-204. ZHU Y N, WANG B, ZHANG C. DNN based signal detection for underwater acoustic FBMC communications[J]. Technical Acoustics, 2021, 40(2): 199-204.
[7]	ZHANG Y, LI J, ZAKHAROV Y, et al. Underwater acoustic OFDM communications using deep learning[C]//The 2^nd Franco-Chinese Acoustic Conference(FCAC). Le Mans, France: Le Mans Université, 2018.
[8]	JIANG R, WANG X, CAO S, et al. Deep neural networks for channel estimation in underwater acoustic OFDM systems[J]. IEEE Access, 2019, 7: 23579-94. doi: 10.1109/ACCESS.2019.2899990
[9]	ZHANG J, CAO Y, HAN G, et al. Deep neural network-based underwater OFDM receiver[J]. IET Communications, 2019, 13(13): 1998-2002. doi: 10.1049/iet-com.2019.0243
[10]	ZHOU M, WANG J, SUN H, et al. A novel DNN based channel estimator for underwater acoustic communications with IM-OFDM[C]//IEEE International Conference on Signal Processing, Communications and Computing (ICSPCC). Macau, China: IEEE, 2020: 1-6.
[11]	HASSAN S, CHEN P, RONG Y, et al. Underwater acoustic OFDM receiver using a regression-based deep neural network[C]//OCEANS 2022. Hampton Roads, VA, USA: IEEE, 2022: 1-6.
[12]	WANG Z, LIU L, CHENG Z, et al. Intelligent receiver design for underwater acoustic OFDM communication based on LSTM networks[C]//IEEE International Conference on Signal Processing, Communications and Computing(ICSPCC). Zhengzhou, China: IEEE, 2023: 1-6.
[13]	HASSAN S, CHEN P, RONG Y, et al. Doppler shift compensation using an LSTM-Based deep neural network in underwater acoustic communication systems[C]//OCEANS 2023. Limerick, Ireland: IEEE, 2023: 1-7.
[14]	ZHANG Y, LI C, WANG H, et al. Deep learning aided OFDM receiver for underwater acoustic communications[J]. Applied Acoustics, 2022, 187: 108515. doi: 10.1016/j.apacoust.2021.108515
[15]	OUYANG D, LI Y, WANG Z. Channel estimation for underwater acoustic OFDM communications: An image super-resolution approach[C]//ICC 2021-IEEE International Conference on Communications. Online: IEEE, 2021: 1-6.
[16]	RUMELHART D E, HINTON G E, WILLIAMS R J. Learning representations by back-propagating errors[J]. Nature, 1986, 323(6088): 533-536. doi: 10.1038/323533a0
[17]	HOCHREITER S. Long short-term memory[J]. Neural Computation, 1997, 9(8): 1735-80. doi: 10.1162/neco.1997.9.8.1735
[18]	QARABAQI P, STOJANOVIC M. Statistical characterization and computationally efficient modeling of a class of underwater acoustic communication channels[J]. IEEE Journal of Oceanic Engineering, 2013, 38(4): 701-717. doi: 10.1109/JOE.2013.2278787