Robust sound event detection in bioacoustic sensor networks

PLoS ONE, Journal Year: 2019, Volume and Issue: 14(10), P. e0214168 - e0214168

Published: Oct. 24, 2019

Bioacoustic sensors, sometimes known as autonomous recording units (ARUs), can record sounds of wildlife over long periods time in scalable and minimally invasive ways. Deriving per-species abundance estimates from these sensors requires detection, classification, quantification animal vocalizations individual acoustic events. Yet, variability ambient noise, both across hinders the reliability current automated systems for sound event detection (SED), such convolutional neural networks (CNN) time-frequency domain. In this article, we develop, benchmark, combine several machine listening techniques to improve generalizability SED models heterogeneous environments. As a case study, consider problem detecting avian flight calls ten-hour nocturnal bird migration, recorded by network six ARUs presence background noise. Starting CNN yielding state-of-the-art accuracy on task, introduce two noise adaptation techniques, respectively integrating short-term (60 ms) long-term (30 min) context. First, apply per-channel energy normalization (PCEN) domain, which applies automatic gain control every subband mel-frequency spectrogram. Secondly, replace last dense layer context-adaptive (CA-NN) layer, i.e. an affine whose weights are dynamically adapted at prediction auxiliary taking summary statistics spectrotemporal features input. We show that PCEN reduces temporal overfitting dawn vs. dusk audio clips whereas context PCEN-based spatial sensor locations. Moreover, combining them yields results unmatched artificial data augmentation alone. release pre-trained version our best performing system under name BirdVoxDetect, ready-to-use detector field recordings.

Language: Английский

---

Measurement artefacts lead to false positives in the study of birdsong in noise

Methods in Ecology and Evolution, Journal Year: 2017, Volume and Issue: 8(11), P. 1617 - 1625

Published: March 10, 2017

Summary Numerous studies over the past decade have reported correlations between elevated levels of anthropogenic noise and a rise in minimum frequency acoustic signals animals living noisy habitats. This pattern appears to be occurring globally, higher pitched been hypothesized adaptive changes that reduce masking by low‐frequency traffic noise. However, sound analysis methods most often used these are prone measurement errors can result false positives. In addition, commonly method measuring frequencies visually from spectrograms might also lead observer‐expectancy biases could exacerbate errors. We conducted an experiment (i) quantify size type ‘eye‐balling’ measurements with cursors placed manually on recorded no‐noise conditions, (ii) test whether observer expectations significant measurements. asked 54 volunteers, blind true intention our study, measure variety natural synthesized bird sounds, either noise, or conditions. Test subjects were informed uninformed about results Our demonstrate inappropriate methodology yield positives effect sizes as large, even larger, than those published studies. addition artefacts, psychological led – when observers expected they measured significantly observers, who had not primed any expectation. The use improper bioacoustics publication spurious results. discuss alternative unbiased measures we caution it is imperative for researchers familiarize themselves both functions limitations their programmes. potential source error only field bioacoustics, but situation where influenced human subjectivity.

Language: Английский

---

Comparative bioacoustics: a roadmap for quantifying and comparing animal sounds across diverse taxa

Biological reviews/Biological reviews of the Cambridge Philosophical Society, Journal Year: 2021, Volume and Issue: 96(4), P. 1135 - 1159

Published: March 2, 2021

Animals produce a wide array of sounds with highly variable acoustic structures. It is possible to understand the causes and consequences this variation across taxa phylogenetic comparative analyses. Acoustic evolutionary analyses are rapidly increasing in sophistication such that choosing appropriate approaches increasingly difficult. However, correct choice analysis can have profound effects on output inferences. Here, we identify address some challenges for growing field by providing roadmap quantifying comparing sound context researchers broad range scientific backgrounds. Sound, as continuous, multidimensional trait be particularly challenging measure because it hard variables compared also no small feat process analyse resulting high-dimensional data using subsequent analysis. Additionally, terminological inconsistencies role learning development traits need considered. Phylogenetic their own sets caveats consider. We provide set recommendations delimiting signals into discrete, comparable units. present three-stage workflow extracting relevant data, including options multivariate dimensionality reduction compatible then summarize available how they been used bioacoustics, limitations behavioural data. Lastly, recommend apply these methods study systems. In way, an integrated framework aid quantitative cross-taxa animal addition, advocate standardization terminology disciplines taxa, adoption automated feature extraction, establishment strong archival practices recordings Combining our proposed will greatly advance reproducibility, biological interpretation, longevity bioacoustic studies.

Language: Английский

---

Soundscapes of urban parks: An innovative approach for ecosystem monitoring and adaptive management

Urban forestry & urban greening, Journal Year: 2022, Volume and Issue: 71, P. 127555 - 127555

Published: April 4, 2022

Language: Английский

---

Weather conditions determine attenuation and speed of sound: Environmental limitations for monitoring and analyzing bat echolocation

Ecology and Evolution, Journal Year: 2018, Volume and Issue: 8(10), P. 5090 - 5100

Published: April 24, 2018

Abstract Echolocating bats are regularly studied to investigate auditory‐guided behaviors and as important bioindicators. Bioacoustic monitoring methods based on echolocation calls increasingly used for risk assessment ultimately inform conservation strategies bats. As transmit through the air at speed of sound, they undergo changes due atmospheric geometric attenuation. Both sound attenuation, however, variable determined by weather conditions, particularly temperature relative humidity. Changing conditions thus cause variation in analyzed call parameters, limiting our ability detect, correctly analyze bat calls. Here, I use real‐world data exemplify effect varying acoustic properties air. then present attenuation global range frequencies show their effects. Atmospheric is a nonlinear function frequency, temperature, humidity, pressure. While strongly positively correlated with it also significantly influenced humidity complex fashion. Variable result unknown effects recorded call, affecting estimates frequency intensity, high frequencies. Weather‐induced reaches up about ±3%, but generally much smaller only relevant localization The frequency‐ weather‐dependent has threefold bioacoustic bats: It limits capability (1) monitor equally across time, space, species, (2) measure parameters calls, frequencies, (3) identify species species‐rich assemblies or sympatric similar designs.

Language: Английский

---

Assessing the effect of sound file compression and background noise on measures of acoustic signal structure

Bioacoustics, Journal Year: 2017, Volume and Issue: 28(1), P. 57 - 73

Published: Nov. 7, 2017

The study of animal acoustic signals is a central tool for many fields in ecology and evolution, but the diversity analytical methods sources sound recordings poses important challenges carrying out robust analyses. Sound file compression background noise may both affect analysis, although little attention has been paid to their respective effects. We evaluated effect these factors by assessing systematic deviation (i.e. bias) measurement error precision) that they generate on spectrographic parameters two (dis)similarity (dynamic time warping frequency contours cross-correlation), which represent most common currently used quantitative characterization signals. Measurements were taken across wide range from diverse group bird species, compared between uncompressed files decompressed obtained mp3-encoded generated using mp3 encoders (Fraunhofer LAME). also synthetically-generated levels. Compression did not significantly bias any or similarity measurements. However, precision representing single extreme values (e.g. peak frequency) as well dynamic distances, was strongly affected compression. High biased energy distribution-related spectral entropy) warping. Overall, have considerable effects provide recommendations avoid potential pitfalls maximize information can be reliably obtained.

Language: Английский

---

Soundscape diversity: Evaluation indices of the sound environment in urban green spaces – Effectiveness, role, and interpretation

Ecological Indicators, Journal Year: 2023, Volume and Issue: 154, P. 110725 - 110725

Published: July 28, 2023

Urban green space sound sources play an important role in providing people with high quality recreation experiences. However, the evaluation index for urban is still limited. The main purpose of this study to propose indices soundscape diversity spaces, compare acoustic diversity, and understand spaces. In our field investigation, 30 sample sites were recorded over a five-minute period, (SDI), richness (SRI), dominance (SDO), evenness (SEI) calculated. results showed that are more suitable than indices. There significant correlation between landscape patterns, there varied relationship perception different Soundscape can be explained by existing objective indicators, which explain 55% SDI model variations, 71% SRI 43% SDO 32% SEI variations. applied future planning environment design.

Language: Английский

---

Aircraft sound exposure leads to song frequency decline and elevated aggression in wild chiffchaffs

Journal of Animal Ecology, Journal Year: 2019, Volume and Issue: 88(11), P. 1720 - 1731

Published: Aug. 21, 2019

The ubiquitous anthropogenic low-frequency noise impedes communication by masking animal signals. To overcome this barrier, animals may increase the frequency, amplitude and delivery rate of their acoustic signals, making them more easily heard. However, a direct impact intermittent, high-level aircraft on birds' behaviour living close to runway has not been studied in detail. We recorded common chiffchaffs Phylloscopus collybita songs near two airports nearby control areas, we measured sound levels territories at Manchester Airport. song recordings were made between movements, when ambient similar airport populations. also conducted playback experiments population test salience airport, specific songs. In contrast general pattern increased frequency noisy show that negative relationship exposure level frequency. Experimental data respond aggressively playback. Since decrease results overlap with noise, these findings cannot be explained as an adaptation improve communication. aggression suggest chiffchaffs, like humans, might affected behaviourally extreme pollution. These should influence environmental assessments for expansions globally.

Language: Английский

---

Using the Past to Understand the Present: Coping with Natural and Anthropogenic Noise

BioScience, Journal Year: 2020, Volume and Issue: 71(3), P. 223 - 234

Published: Dec. 9, 2020

ABSTRACT Anthropogenic noise has received significant attention in recent years, and researchers have highlighted the ways which animals might deal with these sources. However, much of our understanding animal responses to this novel source background acoustics lacks an evolutionary perspective. Natural sources predate origin hearing organs animals. Therefore, it is unlikely that only recently evolved strategies cope anthropogenic de novo but, rather, already preexisting coping mechanisms, because countless generations evolution within a naturally noisy world, on contemporary selection now likely acting. We review natural suggest more quantitative mechanistic how particular characteristics shaped populations communities past, enabling us predict effects will future.

Language: Английский

---

Influence of Anthropogenic Sounds on Insect, Anuran and Bird Acoustic Signals: A Meta-Analysis

Frontiers in Ecology and Evolution, Journal Year: 2022, Volume and Issue: 10

Published: Feb. 28, 2022

Acoustic communication is a way of information exchange between individuals, and it used by several animal species. Therefore, the detection, recognition correct understanding acoustic signals are key factors in effective communication. The priority effectiveness rather than perfection, being avoids affecting sound-based system One that can affect overlap time frequency during signal transmission, known as masking. type sound cause masking anthropogenic noise, which currently increasing due to urban growth consequently motorized transportation machinery. When exposed animals use compensatory mechanisms deal with masking, such modification parameters their signal. Here, we performed meta-analysis investigating whether different taxa have general tendency for changes available literature met minimum criteria perform meta-analysis. We hypothesized noise compensation mechanisms, dominant, maximum or frequencies, call duration, note duration rate meta-analysis, synthesized from 73 studies comprising 82 species three taxa: insects, anurans birds. Our results showed presence insects did not change parameters, while increased amplitude birds dominant frequency, songs. responses groups may be related particularities production reception differences considered also lack some taxa.

Language: Английский

---