CoChat
Sign Up
Cited by 37Underwater Acoustics ResearchBlind Source Separation TechniquesDirection-of-Arrival Estimation Techniques

Passive Sonar Signal Detection and Classification Based on Independent Component Analysis

Open access
de Moura, N. N., de Seixas, J. M., Ramos, Ricardo·2011·Sonar Systems
DOI 10.5772/18286
Open this paper in CoChat — free accountOpen PDF

Abstract

Sonar systems use the sound propagation in underwater environments for detection, communication and navigation. The main purpose of these systems is to analyse the underwater acoustic waves received from different directions by a sensor system and identify the type of target that has been detected in a given direction. Sonar systems may either be

AI snapshot

Written from the paper’s full text.

TL;DR

This chapter applies frequency-domain independent component analysis (ICA) as preprocessing for passive sonar DEMON/LOFAR analysis to reduce interference from neighboring bearings and self-noise, using experimental data from a Brazilian Navy submarine. It reports improved DEMON baselines and mutual information, neural classification recovery to 88% after ICA, and PCA/ART clustering for stationarity monitoring.

Key findings

  • ICA preprocessing reduced baseline noise and interferences for bearings 190° and 205° in DEMON analysis; for bearing 190°, the target shaft peak was at 146.4 rpm, with interfering peaks at 120.1 rpm (205°) and 304.7 rpm (self-noise 076°) before ICA.
  • Mutual information between bearings 190° and 205° was always smaller for ICA-extracted components than for the original mixed signals across the weighting parameter \(p\) from 0 to 1 in steps of 0.01.
  • Neural classification: original unmixed signals gave 94.3%; the text reports adding noise reduces efficiency to 82.7%, mixing signals reduces it further, and applying ICA recovers it to 88%; Table 1 lists without mixer and with noise 89.0%, with mixer 82.7%, and with ICA 88.0%.
  • Stationarity monitoring: PCA retained the first two components, capturing 95% of variance; modified ART clustering on 20 DEMON FFT windows (10 s) detected a statistical change when the cluster count jumped from 3 to 5 between observations 7 and 8.

Methods

  • Experimental raw data were acquired from a passive sonar system installed in a Brazilian Navy submarine and processed with a frequency-domain interference-removal scheme; the ICA block used a Newton-like multiplicative algorithm with kurtosis as cost function and no pre-whitening.
  • DEMON detection used bandpass cavitation filtering, squaring, decimation by 25 twice from 31,250 Hz to 1,250 Hz then DC–50 Hz, STFT, and TPSW; LOFAR used DC–15,625 Hz, a Hanning window, 1,024-sample blocks, STFT, and TPSW.
  • Classification used an MLP with 256 input nodes, one hidden layer of 4 neurons, one output, hyperbolic tangent activation, backpropagation, and half the spectra for training and half for testing, assigning +1 for bearing 31° and -1 for 146°; stationarity analysis used PCA followed by modified ART clustering.

Limitations

  • The paper states passive sonar signals are nonstationary due to ambient noise and external conditions, so stationarity must be monitored and the independent component extraction phase may need reloading/updating when statistics change.
  • The paper states original sources are estimated blindly and suggests other blind source separation methods (e.g., nonnegative matrix factorization, nonstationarity and convolutive mixture algorithms) may improve interference removal.
  • Evident from the text: classification numbers are inconsistent between the narrative and Table 1 (narrative says adding noise reduces to 82.7%, while Table 1 says 89.0%; narrative says mixing further reduces then ICA recovers to 88%, while Table 1 lists with mixer 82.7% and with ICA 88.0%). Also evident: detection improvement is shown qualitatively via DEMON plots and mutual information rather than a formal detection performance metric.

Model-written summary. Check it against the paper before relying on it.

Integrity signals

Registry

Crossref · 10.5772/18286

Cites

30 of 30 references verified0 citation errors · 0 not checkable

Statistics

Nothing to recompute.

Claims the paper makes

1 of 9 cited sources checked · 0 supported · 0 flagged · accuracy not yet measured

  • Premisecitation

    Due to the central limit theorem, the ICA estimates can be obtained by determining the directions for which the nongaussianity is maximum.

    “Due to the well known central limit theorem (Barany, 2007), the ICA estimates can be obtained from the determination of the directions for which the nongaussianity is maximum.”
    Quote in the paperSubstantial claim1 of 1 sources checked
    • cites Barany, 2007 source

      The Abstract reports central limit theorems for the volume and number of faces of Gaussian random polytopes. Section 4 describes the general tendency of normalized sums toward a Gaussian distribution, but neither passage addresses ICA estimates or maximizing nongaussianity over directions. Searches for ICA and for projection directions found no passage supporting that inference. (LLM, conf=0.76)

      Experimental — accuracy not yet measured
  • Findingown data

    From original unmixed signals the neural classification efficiency was 94.3%.

    “From original unmixed signals the classification efficiency was 94.3%.”
    Quote in the paperSubstantial claim
  • Findingown data

    Mixing the bearing signals provides a further reduction in neural classification efficiency.

    “Mixing signals provides a further reduction in classification efficiency”
    Quote in the paperSubstantial claim
  • Findingown data

    Although the components are multiplied by the weighting matrix, the mutual information is always smaller for the ICA-extracted components than for the original mixing signals.

    “Although the components are multiplied by the weighting matrix, the mutual information is always smaller for extracted components than for the original mixing signals.”
    Quote in the paperSubstantial claim
    • The authors state this as an assumption: The source signals in the ICA model are statistically independent.
    • The authors state this as an assumption: The observed signals are linear and instantaneous mixtures of n independent source signals.
  • Findingown data

    Applying ICA recovers the neural classification efficiency, setting it back to 88%.

    “and applying ICA, efficiency is recovered, setting back to 88%.”
    Quote in the paperSubstantial claim
    • The authors state this as an assumption: The source signals in the ICA model are statistically independent.
    • The authors state this as an assumption: The observed signals are linear and instantaneous mixtures of n independent source signals.
  • May combine two claims. 10
  • Classified as probably not a substantial claim. 8

  • May combine two claims.

    Findingown data

    ICA preprocessing reduces the baseline noise for bearings 190 degrees and 205 degrees, improving the signal-to-noise ratio, and reduces interferences from other bearings and the self-noise.

    “In general it is noted that the baseline noise is reduced for bearings 190 o and 205 o , which reflects in a better signal-to-noise ratio.Additionally, interferences from others bearings and the self-noise are reduced.”
    Quote in the paperSubstantial claim
    • The authors state this as an assumption: The source signals in the ICA model are statistically independent.
    • The authors state this as an assumption: The observed signals are linear and instantaneous mixtures of n independent source signals.

  • May combine two claims.

    Findingown data

    For bearing 190 degrees, the largest DEMON peak is at 146.4 rpm, which corresponds to the target shaft rotation.

    “For bearing 190 o the largest peak is at 146.4 rpm, which corresponds to the target shaft rotation.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingown data

    The number of clusters jumped from 3 to 5 between observations 7 and 8, which is a strong indication that the statistics of the signal changed significantly.

    “For example, one can notice that there was a great increase in observation 7 to 8, since the number of clusters jumps from 3 to 5.This is a strong indication that the statistics of the signal changed significantly.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Findingown data

    Adding noise reduces the neural classification efficiency, which reaches 82.7%.

    “Adding noise reduces efficiency, which reaches 82.7%.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Findingown data

    In the DEMON analysis for bearing 190 degrees, interferences are observed at 120.1 rpm and 304.7 rpm, which are relative to the 205 degree bearing and the self-noise (076 degrees), respectively.

    “But interferences are also observed at 120.1 rpm and 304.7 rpm, which are relative to the 205 o and the self-noise (076 o ), respectively.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Findingnone given

    The self-noise at bearing 076 degrees may interfere on the detection of both targets by the sonar operator.

    “Beyond that, the self-noise, bearing 076 • ,may interfere on the detection of both targets by the sonar operator.”
    Quote in the paperNot a substantial claim

  • Classified as probably not a substantial claim. May combine two claims.

    Premisecitation

    The mutual information I(v, w) must be non-negative, and when I(v, w) = 0 the variables v and w are independent.

    “The mutual information I(v, w) must be non-negative.When I(v, w)=0, v and w are independent.”
    Quote in the paperSubstantial claim0 of 2 sources checked
    • cites Hild, 2001

      Couldn’t be matched to a paper in Crossref, so it can’t be checked.

    • cites Hoffmann, 2006

      Couldn’t be matched to a paper in Crossref, so it can’t be checked.

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Premisecitation

    The multiplicative Newton-like ICA algorithm does not require pre-whitening and operates directly over the data.

    “This ICA algorithm does not require pre-whitening and thus operates directly over the data.”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites Akuzawa, 2000

      Couldn’t be matched to a paper in Crossref, so it can’t be checked.

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Premisecitation

    LOFAR is a broadband analysis that provides the machinery noise to the sonar operator and goes from DC to 15,625 Hz.

    “The LOFAR (Waite, 2003) is a broadband analysis that provides the machinery noise to the sonar operator and goes from DC to 15,625 Hz.”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites Waite, 2003

      Couldn’t be matched to a paper in Crossref, so it can’t be checked.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    DEMON is a narrowband analysis that operates over the cavitation noise of the target propeller.

    “DEMON is a narrowband analysis that operates over the cavitation noise of the target propeller (Nielsen, 1991).”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites Nielsen, 1991

      Couldn’t be matched to a paper in Crossref, so it can’t be checked.

  • Classified as probably not a substantial claim.

    Premisecitation

    ICA provides a linear representation of nongaussian data such that the components are statistically independent, or as independent as possible.

    “The ICA provides a linear representation of nongaussian data, so that components are statistically independent, or as much independent as possible (Hyvarinen, 2001), (Yan, 2000).”
    Quote in the paperSubstantial claim0 of 2 sources checked
    • cites Hyvarinen, 2001

      Couldn’t be matched to a paper in Crossref, so it can’t be checked.

    • cites Yan, 2000

      Couldn’t be matched to a paper in Crossref, so it can’t be checked.

  • May combine two claims.

    Interpretationown data

    ICA preprocessing in frequency-domain improves both detection and classification efficiencies, as observed from experimental data.

    “This preprocessing step may be implemented using the independent component analysis in frequency-domain and improvements in both detection and classification efficiencies have been observed from experimental data.”
    Quote in the paperSubstantial claim
    • The authors state this as an assumption: The source signals in the ICA model are statistically independent.
    • The authors state this as an assumption: The observed signals are linear and instantaneous mixtures of n independent source signals.

  • Probably not a substantial claim: may not pass S4 (based).

    Premisenone given

    The passive sonar signals are not stationary due to the ambient noise characteristics and external conditions.

    “The passive sonar signals are not stationary due to the ambient noise characteristics and external conditions.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Study design and conductown data

    The neural classifier training used half of the spectra data from both targets, and classification performance was evaluated on the held-out testing set (the other half of the data that did not participate in training).

    “phase has been performed in batch and used half of the spectra data from both targets.The training algorithm was backpropagation.After training, the classification performance is evaluated from the testing set (the other half of data that did not participate in the training step).”
    Quote in the paperNot a substantial claim
14 of 14

  • The authors state this as an assumption.

    The source signals in the ICA model are statistically independent.

    “with the assumption of independence between the entries of the source vector s(t)”

  • The authors state this as an assumption.

    The observed signals are linear and instantaneous mixtures of n independent source signals.

    “The basic ICA model assumes the existence of n independent signals s 1 (t), ..., s n (t) and the observation of as many linear and instantaneous mixtures x 1 (t), ..., x n (t), as shown in equation 1:”
2 of 2

Self-check

Percentages, sums and reported differences re-computed from the paper’s own numbers.

Each contradiction between two passages listed here was confirmed by an independent check.

  • Noise-only classification efficiency differs from Table 1

    Major

    The text reports 82.7% efficiency after adding noise, but Table 1 assigns 89.0% to the test without mixing and with noise. The table instead assigns 82.7% to the mixed-signal test.

  • Claimed DEMON frequency range exceeds the decimated signal's Nyquist limit

    Minor

    After the first decimation, the sampling rate is 1,250 Hz; a further factor-of-25 decimation makes it 50 Hz. A signal sampled at 50 Hz cannot represent the claimed range up to 50 Hz: its Nyquist limit is 25 Hz.

  • Reported bearings fall outside the stated display range

    Minor

    The bearing display's horizontal axis is stated to run from −180° to 180°, yet the discussion identifies bearings at 190° and 205° on that display.

3 of 3

Cites

Flagged first, then by citations

Research works printed but not linked: 3

  1. Vassali, M. R. ; Seixas, J. M. de and Calôba, L. P.: A Neural Particle Discriminator Based on aM o d i fi e dA r tA r c h i t e c t u r e .IEEE International Symposium on Circuits and Systems, Phoenix, v. II. p. 121-124 (2002).
  2. Yan D. P. and Peach J., Comparison of Blind Souce Separation Algorithms, Advances in Neural Networks and Applications, 18-21 (2000).
  3. Waite A. D., Sonar for practical Engineers, John Wiley, 3rd Ed. (2003). 104 Sonar Systems www.intechopen.com

Cited by

Sorted by citations

Related

Authors
de Moura, N. N., de Seixas, J. M., & Ramos, R.
Affiliations
Universidade Federal do Rio de Janeiro (BR)
Venue
Sonar Systems
Publisher
InTech
Published
2011-09-12
DOI
10.5772/18286
OpenAlex
W1556884246
Semantic Scholar
683a2bb2f5a1839929d745e65bc052041f04b431
Cited by
37
References
28
Citations by year
2025: 2 · 2024: 8 · 2023: 2 · 2022: 3 · 2021: 4 · 2020: 5 · 2019: 4 · 2018: 3 · 2017: 1 · 2016: 2 · 2015: 2 · 2014: 1
FWCI
0.35
Type
book chapter
Language
en
Open access
Licence
cc-by-nc-sa
Locations
  • InTech eBooks (publishedVersion, cc-by-nc-sa)
  • BiblioBoard Library Catalog (Open Research Library) (submittedVersion, cc-by-sa)
  • Contact-less Assessment of In-vivo Body Signals Using Microwave Doppler Radar (InTech) (submittedVersion, cc-by-nc-sa)
Topics
  • Physical Sciences › Earth and Planetary Sciences › Oceanography › Underwater Acoustics Research
  • Physical Sciences › Computer Science › Signal Processing › Blind Source Separation Techniques
  • Physical Sciences › Computer Science › Signal Processing › Direction-of-Arrival Estimation Techniques
Keywords
passive sonar, underwater acoustic signals, independent component analysis, target classification

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

Claims read by deepseek-v4.1-flash under definition version 3 (extract-v3) on 2026-10-17.

Classified by jev-1.13-20260917 (gold-guideline-v0.2) on 2026-10-17.

Cited sources are checked by gpt-6-sol.

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review