Detection and Recognition of Underwater Acoustic Communication Signal Under Ocean Background Noise

Non-cooperative underwater acoustic signal detection and recognition technologies occupy an important position in the marine communication environment. Traditional methods require threshold adjustments based on the environment, making the application less robust. This article addresses the technical...

Full description

Saved in:
Bibliographic Details
Main Authors: Mengyi Li, Jilong Li, Haihong Feng
Format: Article
Language:English
Published: IEEE 2024-01-01
Series:IEEE Access
Subjects:
Online Access:https://ieeexplore.ieee.org/document/10711178/
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1850282334335533056
author Mengyi Li
Jilong Li
Haihong Feng
author_facet Mengyi Li
Jilong Li
Haihong Feng
author_sort Mengyi Li
collection DOAJ
description Non-cooperative underwater acoustic signal detection and recognition technologies occupy an important position in the marine communication environment. Traditional methods require threshold adjustments based on the environment, making the application less robust. This article addresses the technical challenges of automatic detection and recognition of non-cooperative underwater communication signals. It utilizes two pre-trained models, SSD(Single Shot Detector)-MobileNet-V2-FPNlite(Feature Pyramid Network lite) and EfficientDet, which can simultaneously perform signal classification and automatic drawing of signal bounding boxes on the time-frequency graph. SSD-MobileNet-V2-FPNlite uses the Linear Bottleneck layers and the Inverted Residual structure, which improve the capture of multi-scale representations of signals. EfficientDet performs bidirectional feature fusion to enhance the connections and information flow between signal features. After fine-tuning with simulated data, SSD-MobileNet-V2-FPNlite achieves high detection accuracy on both simulated and real lake test data without transfer learning. EfficientDet also performs well on simulated data and, after transfer learning, shows high accuracy in recognizing and drawing bounding boxes on the lake test data. Both methods exhibit strong generalization ability, feature fewer model parameters, and have lower computational complexity.
format Article
id doaj-art-5c7b65a2a34b4da495228374e499eaa1
institution OA Journals
issn 2169-3536
language English
publishDate 2024-01-01
publisher IEEE
record_format Article
series IEEE Access
spelling doaj-art-5c7b65a2a34b4da495228374e499eaa12025-08-20T01:48:02ZengIEEEIEEE Access2169-35362024-01-011214943214944610.1109/ACCESS.2024.347649410711178Detection and Recognition of Underwater Acoustic Communication Signal Under Ocean Background NoiseMengyi Li0https://orcid.org/0009-0000-8599-5999Jilong Li1https://orcid.org/0000-0001-7321-3056Haihong Feng2https://orcid.org/0000-0002-8568-5357Shanghai Acoustics Laboratory, Chinese Academy of Sciences, Shanghai, ChinaShanghai Acoustics Laboratory, Chinese Academy of Sciences, Shanghai, ChinaShanghai Acoustics Laboratory, Chinese Academy of Sciences, Shanghai, ChinaNon-cooperative underwater acoustic signal detection and recognition technologies occupy an important position in the marine communication environment. Traditional methods require threshold adjustments based on the environment, making the application less robust. This article addresses the technical challenges of automatic detection and recognition of non-cooperative underwater communication signals. It utilizes two pre-trained models, SSD(Single Shot Detector)-MobileNet-V2-FPNlite(Feature Pyramid Network lite) and EfficientDet, which can simultaneously perform signal classification and automatic drawing of signal bounding boxes on the time-frequency graph. SSD-MobileNet-V2-FPNlite uses the Linear Bottleneck layers and the Inverted Residual structure, which improve the capture of multi-scale representations of signals. EfficientDet performs bidirectional feature fusion to enhance the connections and information flow between signal features. After fine-tuning with simulated data, SSD-MobileNet-V2-FPNlite achieves high detection accuracy on both simulated and real lake test data without transfer learning. EfficientDet also performs well on simulated data and, after transfer learning, shows high accuracy in recognizing and drawing bounding boxes on the lake test data. Both methods exhibit strong generalization ability, feature fewer model parameters, and have lower computational complexity.https://ieeexplore.ieee.org/document/10711178/Artificial intelligenceacoustic signal detectionacoustic signal processingobject recognitionunderwater acoustics
spellingShingle Mengyi Li
Jilong Li
Haihong Feng
Detection and Recognition of Underwater Acoustic Communication Signal Under Ocean Background Noise
IEEE Access
Artificial intelligence
acoustic signal detection
acoustic signal processing
object recognition
underwater acoustics
title Detection and Recognition of Underwater Acoustic Communication Signal Under Ocean Background Noise
title_full Detection and Recognition of Underwater Acoustic Communication Signal Under Ocean Background Noise
title_fullStr Detection and Recognition of Underwater Acoustic Communication Signal Under Ocean Background Noise
title_full_unstemmed Detection and Recognition of Underwater Acoustic Communication Signal Under Ocean Background Noise
title_short Detection and Recognition of Underwater Acoustic Communication Signal Under Ocean Background Noise
title_sort detection and recognition of underwater acoustic communication signal under ocean background noise
topic Artificial intelligence
acoustic signal detection
acoustic signal processing
object recognition
underwater acoustics
url https://ieeexplore.ieee.org/document/10711178/
work_keys_str_mv AT mengyili detectionandrecognitionofunderwateracousticcommunicationsignalunderoceanbackgroundnoise
AT jilongli detectionandrecognitionofunderwateracousticcommunicationsignalunderoceanbackgroundnoise
AT haihongfeng detectionandrecognitionofunderwateracousticcommunicationsignalunderoceanbackgroundnoise