• Volume/Page
  • Keyword
  • DOI
  • Citation
  • Advanced
   
 
 
 
Journal of the Acoustical Society of America / Volume 131 / Issue 2 / JASA EXPRESS LETTERS

Multifrequency species classification of acoustic-trawl survey data using semi-supervised learning with class discovery

J. Acoust. Soc. Am. Volume 131, Issue 2, pp. EL184-EL190 (2012); (7 pages)

M. Woillez1, P. H. Ressler1, C. D. Wilson1, and J. K. Horne2

1Alaska Fisheries Science Center, National Marine Fisheries Service, NOAA, 7600 Sand Point Way NE, Seattle, Washington 98115 mathieu.woillez@gmail.com, patrick.ressler@noaa.gov, chris.wilson@noaa.gov
2School of Aquatic and Fishery Sciences, University of Washington, Box 355020, Seattle, Washington 98195 jhorne@u.washington.edu

Full Text: Read Online (HTML) | Download PDF FREE | View Cart
Acoustic surveys often use multifrequency backscatter to estimate fish and plankton abundance. Direct samples are used to validate species classification of acoustic backscatter, but samples may be sparse or unavailable. A generalized Gaussian mixture model was developed to classify multifrequency acoustic backscatter when not all species classes are known. The classification, based on semi-supervised learning with class discovery, was applied to data collected in the eastern Bering Sea during summers 2004, 2007, and 2008. Walleye pollock, euphausiids, and two other major classes occurring in the upper water column were identified.

© 2012 Acoustical Society of America

Acknowledgment

Support was provided by Alaska Fishery Science Center, NMFS, NOAA, and the Bering Sea Integrated Ecosystem Research Program (publication #34), North Pacific Research Board (publication #327). We thank Alex De Robertis for supplying labeled data and reviewing the manuscript, Jim Ianelli and two anonymous referees for manuscript review, and Ainhoa Lezama-Ochoa for helpful discussions. Findings and conclusions are those of the authors and do not necessarily represent views of NMFS, NOAA. Reference to trade names does not imply endorsement.

Article Outline

  1. Introduction
  2. Multifrequency acoustic and trawl data processing
  3. Classification procedure
    1. Supervised learning of labeled data
    2. Unsupervised learning of unlabeled data
    3. Semi-supervised learning based on generalized Gaussian mixture model
    4. Treatment of the year effect and model selection
  4. Results and discussion
    1. Classification of multifrequency data
    2. Comparison with other classification methods
  5. Conclusion

RELATED DATABASES

Web of Science

View this record in Web of Science

View Web of Science related articles

KEYWORDS and PACS

Keywords

aquaculture, bioacoustics, Gaussian distribution, learning (artificial intelligence), microorganisms, underwater sound

PACS

  • 43.30.Sf

    Acoustical detection of marine life; passive and active

  • 43.60.Bf

    Acoustic signal detection and classification, applications to control systems

  • 43.60.Lq

    Acoustic imaging, displays, pattern recognition, feature extraction

ARTICLE DATA

History
Received 17 Nov 2011
Accepted 30 Dec 2011
Published online 26 Jan 2012
Digital Object Identifier
http://dx.doi.org/10.1121/1.3678685

PUBLICATION DATA

ISSN

0001-4966 (print)  

Publisher

$abstract.society.value is a Member of CrossRef Acoustical Society of America

  1. J. Simmonds and D. MacLennan, Fisheries Acoustics: Theory and Practice (Blackwell Science, Oxford, 2005).
  2. J. K. Horne, “Acoustic approaches to remote species identification: A review,” Fish. Oceanogr. 9, 356–371 (2000).
  3. P. G. Fernandes, “Classification trees for species identification of fish-school echotraces,” ICES J. Mar. Sci. 66, 1073–1080 (2009).
  4. A. De Robertis, D. R. McKelvey, and P. H. Ressler, “Development and application of an empirical multifrequency method for backscatter classification,” Can. J. Fish. Aquat. Sci. 67, 1459–1474 (2010).
  5. R. J. Korneliussen, Y. Heggelund, I. K. Eliassen, and G. O. Johansen, “Acoustic species identification of schooling fish,” ICES J. Mar. Sci. 66, 1111–1118 (2009).
  6. C. I. H. Anderson, J. K. Horne, and J. Boyle, “Classifying multifrequency fisheries acoustic data using a robust probabilistic classification technique,” J. Acoust. Soc. Am. 121, 230–237 (2007).
  7. K. Nigam, A. McCallum, S. Thrun, and T. Mitchell, “Text classification from labeled and unlabeled documents using EM,” Mach. Learn. 39, 103–134 (2000).
  8. D. J. Miller and J. Browning, “A mixture model and EM-based algorithm for class discovery, robust classification, and outlier rejection in mixed labeled/unlabeled data sets,” IEEE T. Pattern Anal. 25, 1468–1483 (2003).
  9. T. Honkalehto, W. Patton, S. De Blois, and N. J. Williamson, “Echo integration trawl survey results for walleye pollock (Theragra chalcogramma) on the Bering Sea shelf and slope during summer 2000,” NOAA Technical Memorandum No. NMFS AFSC-126 (2002).
  10. A. De Robertis and I. Higginbottom,“A post-processing technique to estimate the signal-to-noise ratio and remove echosounder background noise,” ICES J. Mar. Sci. 64, 1282–1291 (2007).
  11. G. J. McLachlan, Discriminant Analysis and Statistical Pattern Recognition (Wiley, New York, 1992).
  12. A. P. Dempster, N. M. Laird, and D. B. Rubin, “Maximum likelihood from incomplete data via EM algorithm,” J. R. Stat. Soc. Ser. B (Methodol.) 39, 1–38 (1977).
  13. C. Biernacki, G. Celeux, and G. Govaert, “Assessing a mixture model for clustering with the integrated completed likelihood,” IEEE T. Pattern Anal. 22, 719–725 (2000).
  14. M. V. Trevorrow, D. L. Mackas, M. C. Benfield, “Comparison of multifrequency acoustic and in situ measurements of zooplankton abundances in Knight Inlet, British Columbia,” J. Acoust. Soc. Am. 117, 3574–3588 (2005).
  15. P. H. Ressler, A. De Robertis, J. D. Warren, J. N. Smith, and S. Kotwicki, “Developing an acoustic index of euphausiid abundance to understand trophic interactions in the Bering Sea ecosystem,” Deep Sea Res. II (in press).

Figures (3) Tables (1)

Figures (click on thumbnails to view enlargements)

FIG.1
Volume backscattering strength (Sv in dB re 1 m−1) at 18, 38, 120, and 200 kHz (a, b, c, and d) that met signal-to-noise (SNR ≥ 10 dB at any frequency) and threshold (Sv > −80 dB re 1 m−1 at one or more frequencies) criteria from representative transect across the EBS shelf, summer 2007.

FIG.1 Download High Resolution Image (.zip file) | Export Figure to PowerPoint

FIG.2
Example pairwise frequency differences in survey backscatter data for years 2004 (a), 2007 (b), and 2008 (c). 1D and 2D histograms from ΔSv 18-38 and ΔSv 120-38 are represented, as well as the eight fitted multivariate Gaussian class components (colors orange to blue). The fitted mixture model, representing the sum of all class components, is represented by a red line in each 1D histogram. The known class components, pollock, euphausiids, and high18, are indicated concurrently with the unknown class components.

FIG.2 Download High Resolution Image (.zip file) | Export Figure to PowerPoint

FIG.3
Corresponding classes obtained by the classification procedure for a representative transect during summer 2004 (a), 2007 (b), and 2008 (c). Known classes are pollock, euphausiid, and high18. Unknown classes correspond to outliers from a known class, mixtures of known classes, or an undiscovered class.

FIG.3 Download High Resolution Image (.zip file) | Export Figure to PowerPoint

Tables

Table I. Class estimates of alternative classification methods for pollock and euphausiid were compared to corresponding GGMM class estimates by linear regression. Transect mean backscatter was used as the comparative variable. From all possible comparisons (n = 51), only significant (α < 0.001) models with a high r2 and a slope greater than 0.5 are shown. k: for known class component.

View Table


Close

Your Recent History Recent History Help

Recently Viewed

  1. Multifrequency species classification of acoustic-trawl survey data using semi-supervised learning with class discovery
  2. Psychometric properties of the coordinate response measure corpus with various types of background interference
  3. Contrasting behavior between dispersive seismic velocity and attenuation: Advantages in subsoil characterization
  4. Multiuser interference cancellation in time-varying channels
  5. A selective array activation method for the generation of a focused source considering listening position

Home Publications Meetings Contact ASA Site Map Back to Top

Copyright © Acoustical Society of America

close