Characterizing soundscapes across diverse ecosystems using a universal acoustic feature set

Abstract

<jats:p>Natural habitats are being impacted by human pressures at an alarming rate. Monitoring these ecosystem-level changes often requires labor-intensive surveys that are unable to detect rapid or unanticipated environmental changes. Here we have developed a generalizable, data-driven solution to this challenge using eco-acoustic data. We exploited a convolutional neural network to embed soundscapes from a variety of ecosystems into a common acoustic space. In both supervised and unsupervised modes, this allowed us to accurately quantify variation in habitat quality across space and in biodiversity through time. On the scale of seconds, we learned a typical soundscape model that allowed automatic identification of anomalous sounds in playback experiments, providing a potential route for real-time automated detection of irregular environmental behavior including illegal logging and hunting. Our highly generalizable approach, and the common set of features, will enable scientists to unlock previously hidden insights from acoustic data and offers promise as a backbone technology for global collaborative autonomous ecosystem monitoring efforts.</jats:p>

Uniform Manifold Approximation and Projection for Clustering Taxa through Vocalizations in a Neotropical Passerine (Rough-Legged Tyrannulet, Phyllomyias burmeisteri)

“…Such techniques aim to provide an improved classification of bird species with inconclusive scientific evidence to determine the taxonomic status of the species that are difficult to identify [11]. With the advent of machine learning and big data analysis, novel dimensionality reduction techniques have become powerful tools for data analysis enabling a better visualization and understanding of large, high dimensional datasets [12,13]. Dimensionality reduction plays an important role in multiple-dimension data analysis, as it is a fundamental technique for visualization and data processing.…”

Soundscapes predict species occurrence in tropical forests

“…The CNN was trained to perform a general-purpose audio classification task using an extremely large annotated dataset (Gemmeke et al, 2017), resulting in a general 128-dimensional acoustic feature embedding. Prior work has shown that embedding eco-acoustic data using this approach allows multi-scale monitoring of ecosystems and efficient characterisation of soundscapes (Sethi et al, 2020b).…”

“…Analysing soundscapes in their entirety provides an alternate route to the automated analysis of ecoacoustic data (Pijanowski et al, 2011). In this approach, features of the audio signal are used to directly infer habitat quality, without the need for species specific training data (Pieretti et al, 2011;Sethi et al, 2020b;Sueur et al, 2008). Whilst soundscape features have been shown to correlate with high-level metrics of biodiversity, they are not normally used to provide direct evidence for how suitable a habitat is for a given species.…”

Expanding conservation culturomics and iEcology from terrestrial to aquatic realms

“…While most such recordings would not be useful as data sources, the sheer number of videos uploaded daily will ensure that even a very small proportion of usable documents will result in large data sets suitable for analysis. Soundscape assessment approaches have been already demonstrated in the terrestrial realm [ 43 ]. Digital sources could also be mined for past occurrences as well as used for monitoring species in real time.…”