Porpoise Click Classifier (PorCC): a high-accuracy classifier to study harbour porpoises (Phocoena phocoena) in the wild

Research output: Contribution to journalArticle

Abstract

Harbour porpoises are well-suited for passive acoustic monitoring (PAM) as they produce highly stereotyped narrow-band high-frequency (NBHF) echolocation clicks. PAM systems must be coupled with a classification algorithm to identify the signals of interest. Here, we present a harbour porpoise click classifier (PorCC) developed in MATLAB, which uses the coefficients of two logistic regression models in a decision-making pathway to assign candidate signals to one of three categories: high-quality clicks (HQ), low-quality clicks (LQ), or high-frequency noise (N). The receiver operating characteristics of PorCC was compared to that of PAMGuard's Porpoise Click Detector/Classifier Module. PorCC outperformed PAMGuard’s classifier achieving higher hit rates (correctly classified clicks) and lower false alarm levels (noise classified as HQ or LQ clicks). Additionally, the detectability index (d') for HQ clicks for PAMGuard was 2.2 (overall d' = 2.0) versus 4.1 for PorCC (overall d' = 3.4). PorCC classification algorithm is a rapid and highly accurate method to classify NBHF clicks, which could be applied for real time monitoring, as well as to study harbour porpoises, and potentially other NBHF species, throughout their distribution range from data collected using towed hydrophones or static recorders. Moreover, PorCC is suitable for studies of acoustic communication of porpoises.

LanguageEnglish
Number of pages27
JournalJournal of the Acoustical Society of America
Publication statusAccepted/In press - 21 May 2019

Fingerprint

porpoises
harbors
Ports and harbors
classifiers
Classifiers
Acoustics
narrowband
Monitoring
acoustics
Hydrophones
Harbors
Classifier
MATLAB
hydrophones
recorders
false alarms
logistics
Logistics
decision making
Decision making

Keywords

  • classification
  • echolocation
  • logistic regression
  • NBHF species

Cite this

@article{0a04d70d228045928fadcd96f806b57f,
title = "Porpoise Click Classifier (PorCC): a high-accuracy classifier to study harbour porpoises (Phocoena phocoena) in the wild",
abstract = "Harbour porpoises are well-suited for passive acoustic monitoring (PAM) as they produce highly stereotyped narrow-band high-frequency (NBHF) echolocation clicks. PAM systems must be coupled with a classification algorithm to identify the signals of interest. Here, we present a harbour porpoise click classifier (PorCC) developed in MATLAB, which uses the coefficients of two logistic regression models in a decision-making pathway to assign candidate signals to one of three categories: high-quality clicks (HQ), low-quality clicks (LQ), or high-frequency noise (N). The receiver operating characteristics of PorCC was compared to that of PAMGuard's Porpoise Click Detector/Classifier Module. PorCC outperformed PAMGuard’s classifier achieving higher hit rates (correctly classified clicks) and lower false alarm levels (noise classified as HQ or LQ clicks). Additionally, the detectability index (d') for HQ clicks for PAMGuard was 2.2 (overall d' = 2.0) versus 4.1 for PorCC (overall d' = 3.4). PorCC classification algorithm is a rapid and highly accurate method to classify NBHF clicks, which could be applied for real time monitoring, as well as to study harbour porpoises, and potentially other NBHF species, throughout their distribution range from data collected using towed hydrophones or static recorders. Moreover, PorCC is suitable for studies of acoustic communication of porpoises.",
keywords = "classification, echolocation, logistic regression, NBHF species",
author = "Mel Cosentino and Francesco Guarato and Jakob Tougaard and David Nairn and Jackson, {Joseph C.} and Windmill, {James F. C.}",
note = "The following article has been accepted by Journal of the Acoustical Society of America. After it is published, it will be found at https://asa.scitation.org/journal/jas.",
year = "2019",
month = "5",
day = "21",
language = "English",
journal = "Journal of the Acoustical Society of America",
issn = "0001-4966",

}

TY - JOUR

T1 - Porpoise Click Classifier (PorCC)

T2 - Journal of the Acoustical Society of America

AU - Cosentino, Mel

AU - Guarato, Francesco

AU - Tougaard, Jakob

AU - Nairn, David

AU - Jackson, Joseph C.

AU - Windmill, James F. C.

N1 - The following article has been accepted by Journal of the Acoustical Society of America. After it is published, it will be found at https://asa.scitation.org/journal/jas.

PY - 2019/5/21

Y1 - 2019/5/21

N2 - Harbour porpoises are well-suited for passive acoustic monitoring (PAM) as they produce highly stereotyped narrow-band high-frequency (NBHF) echolocation clicks. PAM systems must be coupled with a classification algorithm to identify the signals of interest. Here, we present a harbour porpoise click classifier (PorCC) developed in MATLAB, which uses the coefficients of two logistic regression models in a decision-making pathway to assign candidate signals to one of three categories: high-quality clicks (HQ), low-quality clicks (LQ), or high-frequency noise (N). The receiver operating characteristics of PorCC was compared to that of PAMGuard's Porpoise Click Detector/Classifier Module. PorCC outperformed PAMGuard’s classifier achieving higher hit rates (correctly classified clicks) and lower false alarm levels (noise classified as HQ or LQ clicks). Additionally, the detectability index (d') for HQ clicks for PAMGuard was 2.2 (overall d' = 2.0) versus 4.1 for PorCC (overall d' = 3.4). PorCC classification algorithm is a rapid and highly accurate method to classify NBHF clicks, which could be applied for real time monitoring, as well as to study harbour porpoises, and potentially other NBHF species, throughout their distribution range from data collected using towed hydrophones or static recorders. Moreover, PorCC is suitable for studies of acoustic communication of porpoises.

AB - Harbour porpoises are well-suited for passive acoustic monitoring (PAM) as they produce highly stereotyped narrow-band high-frequency (NBHF) echolocation clicks. PAM systems must be coupled with a classification algorithm to identify the signals of interest. Here, we present a harbour porpoise click classifier (PorCC) developed in MATLAB, which uses the coefficients of two logistic regression models in a decision-making pathway to assign candidate signals to one of three categories: high-quality clicks (HQ), low-quality clicks (LQ), or high-frequency noise (N). The receiver operating characteristics of PorCC was compared to that of PAMGuard's Porpoise Click Detector/Classifier Module. PorCC outperformed PAMGuard’s classifier achieving higher hit rates (correctly classified clicks) and lower false alarm levels (noise classified as HQ or LQ clicks). Additionally, the detectability index (d') for HQ clicks for PAMGuard was 2.2 (overall d' = 2.0) versus 4.1 for PorCC (overall d' = 3.4). PorCC classification algorithm is a rapid and highly accurate method to classify NBHF clicks, which could be applied for real time monitoring, as well as to study harbour porpoises, and potentially other NBHF species, throughout their distribution range from data collected using towed hydrophones or static recorders. Moreover, PorCC is suitable for studies of acoustic communication of porpoises.

KW - classification

KW - echolocation

KW - logistic regression

KW - NBHF species

UR - https://asa.scitation.org/journal/jas

M3 - Article

JO - Journal of the Acoustical Society of America

JF - Journal of the Acoustical Society of America

SN - 0001-4966

ER -