CRCL: Category Restricted Contrastive Learning for Ship-Radiated Noise Recognition Under Mismatched Conditions
IEEE Sensors Journal · 01.07.2025 · 25 (13) · 26117-26131
DOI: 10.1109/jsen.2025.3570232 ↗Kurz gefasst
Fachartikel in der Zeitschrift IEEE Sensors Journal, erschienen am 01.07.2025. Verfasst von Yuanye Zhang und 5 weiteren Autor:innen. Die Arbeit wurde bisher 5-mal zitiert (Crossref) und nennt 60 Literaturangaben. Gefördert von: National Natural Science Foundation of China, National Key Research and Development Program of China, Fundamental Research Funds for the Central Universities.
Abstract (Originalsprache)
The accuracy of ship-radiated noise recognition is often compromised in the complex marine environment due to the influence of environmental mismatch conditions. To address this issue, we simulate a mismatched environmental data and propose a category restricted contrastive learning (CRCL) model for environmental mismatch. To simulate the altered environmental data, we evaluated the performance of the same model across different signal-to-noise ratios (SNRs) and selected an SNR of 15 dB. This choice was based on the observed decrease in recognition performance, while ensuring data distribution remained consistent with the Noise-free dataset. Consequently, we constructed a dataset representative of environmental discrepancies by incorporating the chosen SNR. With mismatched datasets, the CRCL method applies category contrastive loss and a sliding weight approach to identify invariant features, which enhances recognition performance for different types of ship-radiated noise. Ablation experiments demonstrate that the proposed CRCL model significantly improves ship-radiated noise recognition accuracy under various environmental conditions, validating the effectiveness of the proposed strategies.
Förderung
- National Natural Science Foundation of China
- National Key Research and Development Program of China
- Fundamental Research Funds for the Central Universities
Zitieren
APA
Zhang, Y., Gao, L., Zhu, M., Li, N., Li, S., & Huang, Y. (2025). CRCL: Category Restricted Contrastive Learning for Ship-Radiated Noise Recognition Under Mismatched Conditions. IEEE Sensors Journal, 25(13), 26117-26131. https://doi.org/10.1109/jsen.2025.3570232
Vancouver
Zhang Y, Gao L, Zhu M, Li N, Li S, Huang Y. CRCL: Category Restricted Contrastive Learning for Ship-Radiated Noise Recognition Under Mismatched Conditions. IEEE Sensors Journal. 2025;25(13):26117-26131. doi:10.1109/jsen.2025.3570232
BibTeX
@article{Zhang2025,
author = {Zhang, Yuanye and Gao, Lina and Zhu, Minwen and Li, Ning and Li, Shouyang and Huang, Yulong},
title = {CRCL: Category Restricted Contrastive Learning for Ship-Radiated Noise Recognition Under Mismatched Conditions},
journal = {IEEE Sensors Journal},
year = {2025},
volume = {25},
number = {13},
pages = {26117--26131},
doi = {10.1109/jsen.2025.3570232}
}
EUVIMED