Introducing the Software CASE (Cluster and Analyze Sound Events) by Comparing Different Clustering Methods and Audio Transformation Techniques Using Animal Vocalizations

Abstract

Unsupervised clustering algorithms are widely used in ecology and conservation to classify animal sounds, but also offer several advantages in basic bioacoustics research. Consequently, it is important to overcome the existing challenges. A common practice is extracting the acoustic features of vocalizations one-dimensionally, only extracting an average value for a given feature for the entire vocalization. With frequency-modulated vocalizations, whose acoustic features can change over time, this can lead to insufficient characterization. Whether the necessary parameters have been set correctly and the obtained clustering result reliably classifies the vocalizations subsequently often remains unclear. The presented software, CASE, is intended to overcome these challenges. Established and new unsupervised clustering methods (community detection, affinity propagation, HDBSCAN, and fuzzy clustering) are tested in combination with various classifiers (k-nearest neighbor, dynamic time-warping, and cross-correlation) using differently transformed animal vocalizations. These methods are compared with predefined clusters to determine their strengths and weaknesses. In addition, a multidimensional data transformation procedure is presented that better represents the course of multiple acoustic features. The results suggest that, especially with frequency-modulated vocalizations, clustering is more applicable with multidimensional feature extraction compared with one-dimensional feature extraction. The characterization and clustering of vocalizations in multidimensional space offer great potential for future bioacoustic studies. The software CASE includes the developed method of multidimensional feature extraction, as well as all used clustering methods. It allows quickly applying several clustering algorithms to one data set to compare their results and to verify their reliability based on their consistency. Moreover, the software CASE determines the optimal values of most of the necessary parameters automatically. To take advantage of these benefits, the software CASE is provided for free download.

Time series cluster analysis reveals individual assignment of microbiota in captive tiger (Panthera tigris) and wildebeest (Connochaetes taurinus)

“…In this study, we solved the problem using a combined approach of dynamic time warping followed by a clustering algorithm to ensure the comparability of individual time series data and contrast the results with a method using k‐nearest‐neighbor as the classifier (Figure 1 ). The whole clustering pipeline is implemented in Matlab version 9.11 (The MathWorks Inc., 2020 ) using the software CASE (Schneider et al, 2022 ). Accordingly, we applied two different approaches to cluster the data (Figure 2 ).…”

“…Cluster solutions with two or less clusters were ignored except for the species‐specific clustering, as two clusters were to be expected. Furthermore, we used a community detection algorithm (Newman, 2004 ) as implemented in the software CASE (Schneider et al, 2022 ). This agglomerative hierarchical clustering algorithm groups vertices into clusters.…”

DataSheet1.docx

“…Presently, there is no standard in annotating datasets, or which acoustic features should be used in bioacoustics research, when deciding whether two sounds are from the same source (Odom et al, 2021;Schneider et al, 2022). In the present work, AVES and CAE performed differently depending on model configurations and hyperparameters.…”

“…Annotation and labeling, in the absence of a reference dataset with validated annotations, are arduous and subjective especially for an underwater soundscape such as the BPNS, where sound signatures are unknown and the inherent acoustic scene is complex (Parcerisas et al, 2023b). Although unsupervised clustering is conventional in ecological research (Sainburg et al, 2020;Schneider et al, 2022;Guerrero et al, 2023), we highlight its practical use in revising clusters of annotated unknown sounds. Unsupervised clustering and subsequent revision of obtained clusters are therefore proposed steps to systematically reduce annotations to distinct and recurring sound events deemed relevant (Figure 9).…”

“…For example, Ozanich et al (2021) adds an extra clustering layer to an autoencoder similar to the CAE model to penalize points that are distant from cluster centers. If licenses of required MATLAB toolboxes are present, the software CASE (Cluster and Analyze Sound Events; Schneider et al, 2022) can be freely downloaded for the purpose of selecting an appropriate clustering algorithm among four methods (community detection, affinity propagation, HDBSCAN and fuzzy clustering) and three classifiers (k-nearest neighbor, dynamic timewarping and cross-correlation) iterated over different values of parameters. Results are then subsequently evaluated using normalized mutual information (NMI), a scoring metric similar to the homogeneity score which relies on the level of agreement with prelabeled data.…”