针对传统信源编码在水声通信中存在带宽利用率偏低、在低信噪比环境下抗干扰能力不足等问题，本文提出一种基于神经网络的语义信源编码水声通信方法。该方法通过引入Transformer神经网络模块，充分捕捉文本中单词和短语之间的深层次语义关系，生成紧凑和语义丰富的文本表示，大大减少文本信源的冗余，从而在源端实现高效压缩；同时在接收端利用神经网络的强表征能力进行鲁棒解码，以增强系统在复杂信道环境下的稳定性。基于2024年南海实测信道的仿真对比实验表明，相较于传统的信源编码方式，该方法在相同带宽条件下能够显著提升压缩效率和带宽利用率；在低信噪比场景下（–20～–16 dB），有效降低了错词率，通信解码质量提升明显。研究结果表明，基于神经网络的语义信源编码能够为水声通信提供一种兼具高效性与鲁棒性的解决方案，对未来复杂水声信道下的高质量通信具有一定的应用价值。
Traditional source coding in underwater acoustic communications often suffers from low bandwidth efficiency and poor robustness under low signal-to-noise ratio (SNR) conditions. To address these issues, this paper proposes a neural network–based semantic source coding method for underwater acoustic communications. Specifically, a Transformer module is introduced to effectively capture deep semantic relationships between words and phrases, thereby generating compact and semantically rich text representations that greatly reduce redundancy at the source for efficient compression. At the receiver side, the strong representational capacity of neural networks is leveraged to achieve robust decoding, enhancing communication stability in complex channel environments. Simulation experiments based on a measured South China Sea channel in 2024 demonstrate that, compared with traditional source coding schemes, the proposed method significantly improves compression efficiency and bandwidth utilization under the same bandwidth conditions. Moreover, in low-SNR scenarios (-20 dB to -16 dB), it effectively reduces word error rate and achieves notably better decoding quality. These results indicate that the proposed neural network–based semantic source coding provides an efficient and robust solution for underwater acoustic communications, with promising application potential for achieving high-quality communications in future complex underwater channels.
2025,47(24): 141-146 收稿日期：2025-3-12
DOI：10.3404/j.issn.1672-7649.2025.24.022
分类号：U666
作者简介：王戈(1971-),男,硕士,高级工程师,研究方向为水声工程
参考文献：
[1] STOJANOVIC, MILICA. Underwater acoustic communications[C]// In Proceedings of Electro/International, 1995: 435-440.
[2] FARSAD N, RAO M, GOLDSMITH A. Deep learning for joint source-channel coding of text. In 2018 IEEE International Conference on Acoustics[C]//Speech and Signal Processing (ICASSP). IEEE, 2018: 2326-2330.
[3] XIE H, QIN Z, LI G Y, et al. Deep learning enabled semantic communication systems[J]. IEEE Transactions on Signal Processing, 2021(69): 2663-2675.
[4] XIE H, QIN Z. A lite distributed semantic communication system for internet of things[J]. IEEE Journal on Selected Areas in Communications, 2021, 39(1): 142-153.
[5] 张亦弛, 张平, 魏急波, 等. 面向智能体的语义通信: 架构与范例[J]. 中国科学: 信息科学, 2022, 52(5): 907-921.
[6] TANG B, LI Q, HUANG L, et al, Text semantic communication systems with sentence-level semantic fidelity[C]//2023 IEEE Wireless Communications and Networking Conference (WCNC), Glasgow, United Kingdom, 2023: 1-6.
[7] PENG X, QIN Z, TAO X, et al. A robust semantic text communication system[J]. IEEE Transactions on Wireless Communications, 2024, 23(9): 11372-11385.
[8] TAO M, FAN J, LUO J, et al. Coarse-to-fine semantic communication systems for text transmission[J]. IEEE Transactions on Vehicular Technology, 2025, 74(8): 13267-13271.
[9] VASWANI A, SHAZEER N, PARMAR N, et al. Attention is all you need[J]. Advances in Neural Information Processing Systems, 2017: 30.
[10] PAPINENI, KISHORE, SALIM R, et al. Bleu: a method for automatic evaluation of machine translation[C]//In Proceedings of the 40th Annual Meeting of the Association for Computational Linguistics, 2002: 311-318.
