Seabed and range estimation of impulsive time series using a convolutional neural network

Abstract

In ocean acoustics, many types of optimizations have been employed to locate acoustic sources and estimate the properties of the seabed. How these tasks can take advantage of recent advances in deep learning remains as open questions, especially due to the lack of labeled field data. In this work, a Convolutional Neural Network (CNN) is used to find seabed type and source range simultaneously from 1 s pressure time series from impulsive sounds. Simulated data are used to train the CNN before application to signals from a single hydrophone signal during the 2017 Seabed Characterization Experiment. The training data includes four seabeds representing deep mud, mud over sand, sandy silt, and sand, and a wide range of source parameters. When applied to measured data, the trained CNN predicts expected seabed types and obtains ranges within 0.5 km when the source-receiver range is greater than 5 km, showing the potential for such algorithms to address these problems.

A back propagation neural network-based approach for inverting layered seabed acoustic parameters in shallow waters

“…Consequently, establishing a precise functional relationship between acoustic parameters and shallow water acoustic pressure fields presents a significant challenge. To address this challenge, a non-linear mapping approach utilizing a BPNN is employed (Van Komen et al, 2020). On the other hand, the BPNN is capable of approximating functions through the training of input and output vectors.…”

TOSSIT: A low-cost, hand deployable, rope-less and acoustically silent mooring for underwater passive acoustic monitoring

“…Classically, most other ocean acoustics experiments involve the use of synchronized arrays of sensors to perform spatial and temporal filtering. However, recent advances in data science and signal processing now enables localizing a source or characterizing the propagation environment with a single hydrophone [16] , [17] , [18] , [19] , [20] , [21] . Those applications would also greatly benefit from the TOSSIT concept.…”

“…From 2015 to 2017, SBCEX focused on the ‘‘New England Mud Patch’’, a shallow water (depth ∼ 75 m) location about 95 km south of Martha’s Vineyard, MA, USA [8] . Several important SBCEX results were obtained using single receiver studies [18] , [19] , [20] , [21] , which motivated the use for TOSSIT for later experiments. In 2021, the SBCEX study area was extended to cover the deeper water of the New England Shelf Break, about 150 km south of Martha’s Vineyard with water depth up to 500 m. Six TOSSITs were used on the New England Shelf break in 2021 as part of SBCEX, and it is expected that up to 20 TOSSITs will be used in 2022 to cover both the Mud Patch and the Shelf Break.…”

Range Estimation of Few-Shot Underwater Sound Source in Shallow Water Based on Transfer Learning and Residual CNN

“…In 2020, Komen et al input the data obtained by modeling a variety of environmental parameters into the CNN model to realize target range estimation and environmental recognition. The network can effectively estimate the test set obtained by the sound field model, whereas the estimation performance of sea test data was relatively poor [14,15]. Ozanich et al used the FNN for azimuth estimation and compared it with the SVM method.…”