MDJSCC: Multiscale Feature Learning-Based Semantic Communications for Maritime Vessel Networks

Maritime vessel networks face critical communication challenges due to severe signal attenuation induced by dynamic oceanic conditions, including vessel mobility, wave disturbances, and atmospheric variability. These conditions result in high transmission latency and poor communication reliability,...

Celý popis

Uloženo v:
Podrobná bibliografie
Vydáno v:IEEE sensors journal Ročník 25; číslo 22; s. 42251 - 42264
Hlavní autoři: Li, Shibao, Wang, Xinkun, Li, Dongyang, Wang, Chongyang, Meng, Zihan, Zhu, Jinze, Li, Lianghai
Médium: Journal Article
Jazyk:angličtina
Vydáno: New York IEEE 15.11.2025
The Institute of Electrical and Electronics Engineers, Inc. (IEEE)
Témata:
Adaptation models
Communication
Computer architecture
Data exchange
Deep learning
Encoding
Feature extraction
Joint source–channel coding
marine communication
Maritime communications
maritime vessel networks
Multilayers
Networks
Reliability
Sea vessels
semantic
Semantic communication
Semantics
Sensors
Signal to noise ratio
Wireless communication
ISSN:1530-437X, 1558-1748
On-line přístup:Získat plný text
Tagy: Přidat tag
Žádné tagy, Buďte první, kdo vytvoří štítek k tomuto záznamu!
Popis
Shrnutí:Maritime vessel networks face critical communication challenges due to severe signal attenuation induced by dynamic oceanic conditions, including vessel mobility, wave disturbances, and atmospheric variability. These conditions result in high transmission latency and poor communication reliability, making traditional bit-level communication techniques insufficient for supporting efficient and stable data exchange. To~address this issue, this work proposes multiscale feature learning (ML)-driven semantic communications for maritime vessel networks. Specifically, a maritime vessel communication model is constructed with consideration for flexibility and scalability in maritime communications, and the propagation characteristics of marine signals were simulated based on the Rician fading channel. Building upon this model, a~novel semantic communication framework named multiscale deep joint source-channel coding (MDJSCC) is developed. Combining deep learning techniques, MDJSCC enables adaptive transmission strategies based on varying channel states and compression requirements, thereby improving robustness in complex maritime environments. More importantly, a~multiscale learning block is introduced to enhance the feature extraction capability for maritime data by integrating multilayer convolutional networks. Extensive experiments are conducted on the public ocean buoy and marine image datasets. The results demonstrate that the proposed MDJSCC can achieve significant gains over traditional marine communication at low signal-to-noise ratios (SNRs) and outperforms state-of-the-art methods, yielding 22.5% and 12.4% PSNR improvements over DeepJSCC and SwinJSCC while concurrently reducing transmission volume and ensuring robust marine communication performance.
Bibliografie:ObjectType-Article-1
SourceType-Scholarly Journals-1
ObjectType-Feature-2
content type line 14
ISSN:1530-437X
1558-1748
DOI:10.1109/JSEN.2025.3614462