The University of Southampton
University of Southampton Institutional Repository

Parametric modelling of cetacean calls

Parametric modelling of cetacean calls
Parametric modelling of cetacean calls

This thesis is concerned with the detection and characterisation of tonal cetacean call through notch filter modelling.  The recognition of these calls is based on their frequency evolution, and different approaches to estimating this are described.  Adaptive notch filters (ANF) are viewed and compared through simulations.  A novel RLS-based algorithm for estimating the local sweep rate of a frequency evolution is then developed.  Biasing the frequency tracker’s update law towards this sweep rate permits one to resolve confusion problems at frequency crossings as well as during strong interfering sounds, such as cetacean clicks, that would otherwise disturb the tracking.  A click pre-detection algorithm is developed and coupled to the frequency tracker.  Further, a novel detection methodology based on the local reliability of the ANF parameter estimates is developed and analysed. The reliability is measured by the curvature of the cost function Hessian, and detections are made from periods of reliable estimation.

The thesis also investigates block-based approaches to frequency estimation.  Time-varying AR (TVAR) models are improved and their properties in additive noise analysed.  Block-based notch filters are developed and found to be more noise robust than TVAR models.  However, estimation is more difficult than with ANF and the noise robustness is similar.

Comparing the most suitable of the parametric model approaches to a recently published spectrogram-based whistle processing algorithm, it is found that the ANF-based methods are more reliable and have better performance on recordings of tonals and isolated clicks.  They are able to track rapidly sweeping tonals and cope well with frequency crossings.  The detection methodology gives very few false alarms and permits accurate determination of start and end times of multiple simultaneous tonals.

University of Southampton
Johannson, Anders Torbjörn
ebca512a-a8df-45ae-88e3-43043dfe230c
Johannson, Anders Torbjörn
ebca512a-a8df-45ae-88e3-43043dfe230c

Johannson, Anders Torbjörn (2004) Parametric modelling of cetacean calls. University of Southampton, Doctoral Thesis.

Record type: Thesis (Doctoral)

Abstract

This thesis is concerned with the detection and characterisation of tonal cetacean call through notch filter modelling.  The recognition of these calls is based on their frequency evolution, and different approaches to estimating this are described.  Adaptive notch filters (ANF) are viewed and compared through simulations.  A novel RLS-based algorithm for estimating the local sweep rate of a frequency evolution is then developed.  Biasing the frequency tracker’s update law towards this sweep rate permits one to resolve confusion problems at frequency crossings as well as during strong interfering sounds, such as cetacean clicks, that would otherwise disturb the tracking.  A click pre-detection algorithm is developed and coupled to the frequency tracker.  Further, a novel detection methodology based on the local reliability of the ANF parameter estimates is developed and analysed. The reliability is measured by the curvature of the cost function Hessian, and detections are made from periods of reliable estimation.

The thesis also investigates block-based approaches to frequency estimation.  Time-varying AR (TVAR) models are improved and their properties in additive noise analysed.  Block-based notch filters are developed and found to be more noise robust than TVAR models.  However, estimation is more difficult than with ANF and the noise robustness is similar.

Comparing the most suitable of the parametric model approaches to a recently published spectrogram-based whistle processing algorithm, it is found that the ANF-based methods are more reliable and have better performance on recordings of tonals and isolated clicks.  They are able to track rapidly sweeping tonals and cope well with frequency crossings.  The detection methodology gives very few false alarms and permits accurate determination of start and end times of multiple simultaneous tonals.

Text
945103.pdf - Version of Record
Available under License University of Southampton Thesis Licence.
Download (18MB)

More information

Published date: 2004

Identifiers

Local EPrints ID: 465356
URI: http://eprints.soton.ac.uk/id/eprint/465356
PURE UUID: 8164bed8-7866-42ee-a4a7-f5c7e76f5554

Catalogue record

Date deposited: 05 Jul 2022 00:39
Last modified: 16 Mar 2024 20:07

Export record

Contributors

Author: Anders Torbjörn Johannson

Download statistics

Downloads from ePrints over the past year. Other digital versions may also be available to download e.g. from the publisher's website.

View more statistics

Atom RSS 1.0 RSS 2.0

Contact ePrints Soton: eprints@soton.ac.uk

ePrints Soton supports OAI 2.0 with a base URL of http://eprints.soton.ac.uk/cgi/oai2

This repository has been built using EPrints software, developed at the University of Southampton, but available to everyone to use.

We use cookies to ensure that we give you the best experience on our website. If you continue without changing your settings, we will assume that you are happy to receive cookies on the University of Southampton website.

×