Detection and Classification Methods for Animal Sounds

Abstract

Classification of the acoustic repertoires of animals into sound types is a useful tool for taxonomic studies, behavioral studies, and for documenting the occurrence of animals. Classification of acoustic repertoires enables the identification of species, age, gender, and individual identity, correlations between sound types and behavior, the identification of changes in vocal behavior over time or in response to anthropogenic noise, comparisons between the repertoires of populations living in different geographic regions and environments, and the development of software tools for automated signal processing. Techniques for classification have evolved over time as technical capabilities have expanded. Initially, researchers applied qualitative methods, such as listening and visually discerning sounds in spectrograms. Advances in computer technology and the development of software for the automatic detection and classification of sounds have allowed bioacousticians to quickly find sounds in recordings, thus significantly reducing analysis time and enabling the analysis of larger datasets. In this chapter, we present software algorithms for automated signal detection (based on energy, Teager–Kaiser energy, spectral entropy, matched filtering, and spectrogram cross-correlation) as well as for signal classification (e.g., parametric clustering, principal component analysis, discriminant function analysis, classification trees, artificial neural networks, random forests, Gaussian mixture models, support vector machines, dynamic time-warping, and hidden Markov models). Methods for evaluating the performance of automated tools are presented (i.e., receiver operating characteristics and precision-recall) and challenges with classifying animal sounds are discussed.

Call type repertoire of killer whales (Orcinus orca) in Iceland and its variation across regions

“…However, agreement was higher when differentiating at the call type level (i.e., distinguishing different call types or identifying calls of the same type and subtype) but lower on finer variations (i.e., when to differentiate between subtypes). These differences likely arose because some people are "splitters" and others "lumpers" and thus even experienced observers classify sounds differently (Oswald et al, 2022). A shortcoming of the interobserver reliability test was that observers were asked to directly compare between only two call samples, thus lacking information for finer distinctions (i.e., subtypes), such as whether variation is graded or discrete and whether there are intermediates.…”

Bioacoustic monitoring to determine addiction levels of primates to the human sphere: A feasibility study on Japanese macaques

“…To detect targeted calls among the recorded data, tonal information should first be converted into a spectrogram using the Fourier transform (Erbe et al, 2022). To facilitate the automatic detection and classification of targeted calls, various advanced statistical procedures based on machine learning or deep learning techniques are now available (Oswald et al, 2022; Shonfield & Bayne, 2017; Stowell et al, 2019; Sugai et al, 2019; Wu et al, 2022). More recently, there is also an outstanding approach using transfer learning, which could achieve PAM under conditions with limited training data through the use of a model trained on a given data set to predict a related task (examples targeted at primate calls, Dufourq et al, 2022; Ravaglia et al, 2023).…”

“…More recently, there is also an outstanding approach using transfer learning, which could achieve PAM under conditions with limited training data through the use of a model trained on a given data set to predict a related task (examples targeted at primate calls, Dufourq et al, 2022; Ravaglia et al, 2023). Software packages for these analytical supports are readily accessible, including shareware and opensource R packages (for a detailed software list, see Oswald et al, 2022).…”

“…The theory and techniques of PAM have been developed for sound‐producing animals that are difficult to observe directly (e.g., cetaceans, bats, birds, and anurans) since the 1990s (Marques et al, 2013; Sugai et al, 2019). In the last decade, however, the growing use of PAM, even for species that are detectable through the use of visual cues (e.g., camera traps) such as nonflying terrestrial mammals, has been confirmed according to the following shared merits of using sound cues: (1) greater robustness in landscapes with limited visibility (such as dense forests); (2) broader areas to detect target species; and (3) a simpler algorithm for automated species discrimination, compared to visual cues (Marques et al, 2013; Oswald et al, 2022; Sugai et al, 2019). PAM has now been applied for wild primates to assess species distribution (e.g., Macaca , Enari et al, 2019; Leontopithecus , Zambolli et al, 2023), population abundance or density (e.g., Ateles , Hutschenreiter et al, 2022; Phaner , Markolf et al, 2022), species discrimination (e.g., Cercopithecidae and Pan , Heinicke et al, 2015), individual recognition (e.g., Hylobates , Clink & Klinck, 2021), site occupancy (e.g., Cercopithecidae and Pan , Kalan et al, 2015), and diel or seasonal patterns of vocal communications (e.g., Atelidae , Do Nascimento et al, 2021; Pérez‐Granados & Schuchmann, 2021).…”

“…(3) a simpler algorithm for automated species discrimination, compared to visual cues (Marques et al, 2013;Oswald et al, 2022;Sugai et al, 2019). PAM has now been applied for wild primates to assess species distribution (e.g., Macaca, Enari et al, 2019;Leontopithecus, Zambolli et al, 2023), population abundance or density (e.g., Ateles, Hutschenreiter et al, 2022;Phaner, Markolf et al, 2022), species discrimination (e.g., Cercopithecidae and Pan, Heinicke et al, 2015), individual recognition (e.g., Hylobates, Clink & Klinck, 2021), site occupancy (e.g., Cercopithecidae and Pan, Kalan et al, 2015), and diel or seasonal patterns of vocal communications (e.g., Atelidae, Do Nascimento et al, 2021;Pérez-Granados & Schuchmann, 2021).…”