Interpretable and Robust Machine Learning for Exploring and Classifying Soundscape Data DOI Creative Commons
Arpit Omprakash, Rohini Balakrishnan, Robert M. Ewers

et al.

bioRxiv (Cold Spring Harbor Laboratory), Journal Year: 2024, Volume and Issue: unknown

Published: Nov. 8, 2024

Abstract The adoption of machine learning in Passive Acoustic Monitoring (PAM) has improved prediction accuracy for tasks like species-specific call detection and habitat quality estimation. However, these models often lack interpretability, PAM generates vast amounts non-informative data, as soundscapes are typically information sparse. Here, we developed ecologically interpretable methods that accurately predict land use from audio while filtering unwanted data. Audio habitats Southern India (evergreen forests, deciduous scrublands, grasslands) was collected categorised by (reference, disturbed, agriculture). We used Gaussian Mixture Models (GMMs) on top a Convolutional Neural Network (CNN)-based feature extractor to use. Thresholding based likelihood values GMMs model excluding uninformative enabling our method outperform such Random Forests Support Vector Machines. By analysing areas acoustic space driving predictions, identified “keystone” soundscape elements each use, including both biotic anthropogenic sources. Our approach provides novel meaningful interpretation exploration large datasets independent specific extractors. study paves the way monitoring deliver robust trustworthy assessments scales would not otherwise be possible.

Language: Английский

Urban Blue-Green Spaces and tranquility: a comprehensive review of noise reduction and sensory perception integration DOI Creative Commons

S. N. G. Chu,

Weizhen Xu,

Dan-Yin Zhang

et al.

Journal of Asian Architecture and Building Engineering, Journal Year: 2025, Volume and Issue: unknown, P. 1 - 22

Published: March 19, 2025

Language: Английский

Citations

0
Busy Urban Soundscape Underwater: Acoustic Indicators vs. Hydrophone Data DOI Creative Commons

Kamil Monastyrski,

Grzegorz Chrobak, Rengin Aslanoğlu

et al.

Urban Science, Journal Year: 2025, Volume and Issue: 9(4), P. 129 - 129

Published: April 17, 2025

Urban noise pollution extends into aquatic environments, influencing underwater ecosystems. This study examines the effectiveness of acoustic indicators in characterizing urban soundscapes using hydrophone recordings. Three indices, Acoustic Complexity Index (ACI), Diversity (ADI), and Normalized Difference Soundscape (NDSI), were analyzed to assess their ability distinguish anthropogenic natural sources. The results indicate that ACI tracks fluctuations, particularly from vehicles trams, while ADI primarily reflects transient environmental interferences. NDSI, designed differentiate biophony noise, proves unreliable settings, often misclassifying These findings highlight limitations traditional indices environments emphasize need for refined methods improve data interpretation. Thus, this aims understand indicators’ interactions with which is crucial enhancing monitoring mitigation strategies.

Language: Английский

Citations

0
Interpretable and Robust Machine Learning for Exploring and Classifying Soundscape Data DOI Creative Commons
Arpit Omprakash, Rohini Balakrishnan, Robert M. Ewers

et al.

bioRxiv (Cold Spring Harbor Laboratory), Journal Year: 2024, Volume and Issue: unknown

Published: Nov. 8, 2024

Abstract The adoption of machine learning in Passive Acoustic Monitoring (PAM) has improved prediction accuracy for tasks like species-specific call detection and habitat quality estimation. However, these models often lack interpretability, PAM generates vast amounts non-informative data, as soundscapes are typically information sparse. Here, we developed ecologically interpretable methods that accurately predict land use from audio while filtering unwanted data. Audio habitats Southern India (evergreen forests, deciduous scrublands, grasslands) was collected categorised by (reference, disturbed, agriculture). We used Gaussian Mixture Models (GMMs) on top a Convolutional Neural Network (CNN)-based feature extractor to use. Thresholding based likelihood values GMMs model excluding uninformative enabling our method outperform such Random Forests Support Vector Machines. By analysing areas acoustic space driving predictions, identified “keystone” soundscape elements each use, including both biotic anthropogenic sources. Our approach provides novel meaningful interpretation exploration large datasets independent specific extractors. study paves the way monitoring deliver robust trustworthy assessments scales would not otherwise be possible.

Language: Английский

Citations

0