Spectrograms and scalograms with correlation filters for anuran vocalization classification

Salina Abdul Samad, Aqilah Baseri Huddin

Research output: Contribution to journalArticle

Abstract

Correlation filters have been used with spectrograms to classify animals based on their vocalizations. The training images are a series of the animal vocalizations converted into spectrograms with each class having its own corresponding template. Cross-correlation is then performed with the test spectrogram images in order to classify them to the corresponding classes based on the correlation output parameters. In this paper, a similar approach is used using not only spectrograms, but also scalograms for comparisons in classifying anuran vocalization. The results show that although increasing the number of vocalization when constructing the templates increases the accuracy rate for both spectrograms and scalograms, spectrograms are much more suitable to be used for anuran vocalization classification due a better overall performance.

Original languageEnglish
Pages (from-to)594-597
Number of pages4
JournalInternational Journal of Advanced Trends in Computer Science and Engineering
Volume8
Issue number3
DOIs
Publication statusPublished - 1 May 2019

Fingerprint

Animals

Keywords

  • Anuran
  • Classification
  • Correlation filter
  • Scalogram
  • Spectrogram

ASJC Scopus subject areas

  • Computer Science (miscellaneous)
  • Electrical and Electronic Engineering

Cite this

@article{35028cc0d4ea487e9028fdc0e1e821e6,
title = "Spectrograms and scalograms with correlation filters for anuran vocalization classification",
abstract = "Correlation filters have been used with spectrograms to classify animals based on their vocalizations. The training images are a series of the animal vocalizations converted into spectrograms with each class having its own corresponding template. Cross-correlation is then performed with the test spectrogram images in order to classify them to the corresponding classes based on the correlation output parameters. In this paper, a similar approach is used using not only spectrograms, but also scalograms for comparisons in classifying anuran vocalization. The results show that although increasing the number of vocalization when constructing the templates increases the accuracy rate for both spectrograms and scalograms, spectrograms are much more suitable to be used for anuran vocalization classification due a better overall performance.",
keywords = "Anuran, Classification, Correlation filter, Scalogram, Spectrogram",
author = "{Abdul Samad}, Salina and Huddin, {Aqilah Baseri}",
year = "2019",
month = "5",
day = "1",
doi = "10.30534/ijatcse/2019/40832019",
language = "English",
volume = "8",
pages = "594--597",
journal = "International Journal of Advanced Trends in Computer Science and Engineering",
issn = "2278-3091",
publisher = "World Academy of Research in Science and Engineering",
number = "3",

}

TY - JOUR

T1 - Spectrograms and scalograms with correlation filters for anuran vocalization classification

AU - Abdul Samad, Salina

AU - Huddin, Aqilah Baseri

PY - 2019/5/1

Y1 - 2019/5/1

N2 - Correlation filters have been used with spectrograms to classify animals based on their vocalizations. The training images are a series of the animal vocalizations converted into spectrograms with each class having its own corresponding template. Cross-correlation is then performed with the test spectrogram images in order to classify them to the corresponding classes based on the correlation output parameters. In this paper, a similar approach is used using not only spectrograms, but also scalograms for comparisons in classifying anuran vocalization. The results show that although increasing the number of vocalization when constructing the templates increases the accuracy rate for both spectrograms and scalograms, spectrograms are much more suitable to be used for anuran vocalization classification due a better overall performance.

AB - Correlation filters have been used with spectrograms to classify animals based on their vocalizations. The training images are a series of the animal vocalizations converted into spectrograms with each class having its own corresponding template. Cross-correlation is then performed with the test spectrogram images in order to classify them to the corresponding classes based on the correlation output parameters. In this paper, a similar approach is used using not only spectrograms, but also scalograms for comparisons in classifying anuran vocalization. The results show that although increasing the number of vocalization when constructing the templates increases the accuracy rate for both spectrograms and scalograms, spectrograms are much more suitable to be used for anuran vocalization classification due a better overall performance.

KW - Anuran

KW - Classification

KW - Correlation filter

KW - Scalogram

KW - Spectrogram

UR - http://www.scopus.com/inward/record.url?scp=85068773509&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85068773509&partnerID=8YFLogxK

U2 - 10.30534/ijatcse/2019/40832019

DO - 10.30534/ijatcse/2019/40832019

M3 - Article

VL - 8

SP - 594

EP - 597

JO - International Journal of Advanced Trends in Computer Science and Engineering

JF - International Journal of Advanced Trends in Computer Science and Engineering

SN - 2278-3091

IS - 3

ER -