Convolutional neural network with second-order pooling for underwater target classification

Xu Cao, Roberto Togneri, Xiaomin Zhang, Yang Yu

Research output: Contribution to journalArticle

2 Citations (Scopus)

Abstract

Underwater target classification using passive sonar remains a critical issue due to the changeable ocean environment. Convolutional Neural Networks (CNNs) have shown success in learning invariant features using local filtering and max pooling. In this paper, we propose a novel classification framework which combines the CNN architecture with the second-order pooling (SOP) to capture the temporal correlations from the timefrequency (T-F) representation of the radiated acoustic signals. The convolutional layers are used to learn the local features with a set of kernel filters from the T-F inputs which are extracted by the constant-Q transform (CQT). Instead of using max pooling, the proposed SOP operator is designed to learn the co-occurrences of different CNN filters using the temporal feature trajectory of CNN features for each frequency subband. To preserve the frequency distinctions, the correlated features of each frequency subband are retained. The pooling results are normalized with signed square-root and l2 normalization, and then input into the softmax classifier. The whole network can be trained in an end-to-end fashion. To explore the generalization ability to unseen conditions, the proposed CNN model is evaluated on the real radiated acoustic signals recorded at new sea depths. The experimental results demonstrate that the proposed method yields an 8% improvement in classification accuracy over the state-of-the-art deep learning methods.

Original languageEnglish
Pages (from-to)3058-3066
JournalIEEE Sensors Journal
Volume19
Issue number8
Early online date12 Dec 2018
DOIs
Publication statusPublished - 15 Apr 2019

Fingerprint

Neural networks
learning
Acoustics
Sonar
filters
Network architecture
acoustics
sonar
classifiers
Classifiers
Trajectories
oceans
trajectories
occurrences
operators

Cite this

@article{5f274f4769434534a1b518a1cae87731,
title = "Convolutional neural network with second-order pooling for underwater target classification",
abstract = "Underwater target classification using passive sonar remains a critical issue due to the changeable ocean environment. Convolutional Neural Networks (CNNs) have shown success in learning invariant features using local filtering and max pooling. In this paper, we propose a novel classification framework which combines the CNN architecture with the second-order pooling (SOP) to capture the temporal correlations from the timefrequency (T-F) representation of the radiated acoustic signals. The convolutional layers are used to learn the local features with a set of kernel filters from the T-F inputs which are extracted by the constant-Q transform (CQT). Instead of using max pooling, the proposed SOP operator is designed to learn the co-occurrences of different CNN filters using the temporal feature trajectory of CNN features for each frequency subband. To preserve the frequency distinctions, the correlated features of each frequency subband are retained. The pooling results are normalized with signed square-root and l2 normalization, and then input into the softmax classifier. The whole network can be trained in an end-to-end fashion. To explore the generalization ability to unseen conditions, the proposed CNN model is evaluated on the real radiated acoustic signals recorded at new sea depths. The experimental results demonstrate that the proposed method yields an 8{\%} improvement in classification accuracy over the state-of-the-art deep learning methods.",
keywords = "Acoustics, Computer architecture, constant-Q transform, Convolutional neural networks, convolutional neural networks, Correlation, Feature extraction, second-order pooling, Time-frequency analysis, Transforms, underwater target classification",
author = "Xu Cao and Roberto Togneri and Xiaomin Zhang and Yang Yu",
year = "2019",
month = "4",
day = "15",
doi = "10.1109/JSEN.2018.2886368",
language = "English",
volume = "19",
pages = "3058--3066",
journal = "IEEE Sensors Journal",
issn = "1530-437X",
publisher = "IEEE, Institute of Electrical and Electronics Engineers",
number = "8",

}

Convolutional neural network with second-order pooling for underwater target classification. / Cao, Xu; Togneri, Roberto; Zhang, Xiaomin; Yu, Yang.

In: IEEE Sensors Journal, Vol. 19, No. 8, 15.04.2019, p. 3058-3066.

Research output: Contribution to journalArticle

TY - JOUR

T1 - Convolutional neural network with second-order pooling for underwater target classification

AU - Cao, Xu

AU - Togneri, Roberto

AU - Zhang, Xiaomin

AU - Yu, Yang

PY - 2019/4/15

Y1 - 2019/4/15

N2 - Underwater target classification using passive sonar remains a critical issue due to the changeable ocean environment. Convolutional Neural Networks (CNNs) have shown success in learning invariant features using local filtering and max pooling. In this paper, we propose a novel classification framework which combines the CNN architecture with the second-order pooling (SOP) to capture the temporal correlations from the timefrequency (T-F) representation of the radiated acoustic signals. The convolutional layers are used to learn the local features with a set of kernel filters from the T-F inputs which are extracted by the constant-Q transform (CQT). Instead of using max pooling, the proposed SOP operator is designed to learn the co-occurrences of different CNN filters using the temporal feature trajectory of CNN features for each frequency subband. To preserve the frequency distinctions, the correlated features of each frequency subband are retained. The pooling results are normalized with signed square-root and l2 normalization, and then input into the softmax classifier. The whole network can be trained in an end-to-end fashion. To explore the generalization ability to unseen conditions, the proposed CNN model is evaluated on the real radiated acoustic signals recorded at new sea depths. The experimental results demonstrate that the proposed method yields an 8% improvement in classification accuracy over the state-of-the-art deep learning methods.

AB - Underwater target classification using passive sonar remains a critical issue due to the changeable ocean environment. Convolutional Neural Networks (CNNs) have shown success in learning invariant features using local filtering and max pooling. In this paper, we propose a novel classification framework which combines the CNN architecture with the second-order pooling (SOP) to capture the temporal correlations from the timefrequency (T-F) representation of the radiated acoustic signals. The convolutional layers are used to learn the local features with a set of kernel filters from the T-F inputs which are extracted by the constant-Q transform (CQT). Instead of using max pooling, the proposed SOP operator is designed to learn the co-occurrences of different CNN filters using the temporal feature trajectory of CNN features for each frequency subband. To preserve the frequency distinctions, the correlated features of each frequency subband are retained. The pooling results are normalized with signed square-root and l2 normalization, and then input into the softmax classifier. The whole network can be trained in an end-to-end fashion. To explore the generalization ability to unseen conditions, the proposed CNN model is evaluated on the real radiated acoustic signals recorded at new sea depths. The experimental results demonstrate that the proposed method yields an 8% improvement in classification accuracy over the state-of-the-art deep learning methods.

KW - Acoustics

KW - Computer architecture

KW - constant-Q transform

KW - Convolutional neural networks

KW - convolutional neural networks

KW - Correlation

KW - Feature extraction

KW - second-order pooling

KW - Time-frequency analysis

KW - Transforms

KW - underwater target classification

UR - http://www.scopus.com/inward/record.url?scp=85058627798&partnerID=8YFLogxK

U2 - 10.1109/JSEN.2018.2886368

DO - 10.1109/JSEN.2018.2886368

M3 - Article

VL - 19

SP - 3058

EP - 3066

JO - IEEE Sensors Journal

JF - IEEE Sensors Journal

SN - 1530-437X

IS - 8

ER -