๐Ÿ“–
Sound-Based Bird Localization Using Lightweight Convolutional Neural Networks and Internet of Things
Chakir Bouarouguene, Moufida Maimour, Ouahab Kadri, Abderrazak Benyahia, Eric Rondeau

Sound-Based Bird Localization Using Lightweight Convolutional Neural Networks and Internet of Things

Science & Mathematics Published 2025 Verification Pending

Overview & Synopsis

Acoustic bird localization provides valuable insights for environmental health monitoring, which plays a crucial role in green engineering. While Convolutional Neural Networks (CNNs) have shown promise in sound source localization, their application to bird localization faces two main challenges : the difficulty of collecting training data and the resource constraints of Inernet of Things (IoT) devices, including limitations in energy and processing power. To address the first challenge, we generate a high-quality dataset through simulation. The data generation process begins by framing the localization task as a classification problem, followed by a well-distributed placement of sound source points. Realistic environmental factors such as temperature, foliage, and actual bird sounds are then incorporated to create a more representative dataset. Next, we evaluate several existing CNNs for audio classification to identify the most suitable model for our data, which is subsequently compressed to mitigate resource limitations. Our results show that our model achieves an accuracy of 95.59%, representing a 2.21% improvement over state-of-the-art CNNs. Additionally, it reduces model size by 89.44% relative to the VGGish CNN, enabling real-time inference on Fog devices with reduced resource requirements.

Indexed from repository: HAL Open Science (hal.science)
Loading document from verified source...
Keyboard Shortcuts: ← / → for pages • + / - for zoom • 0 to fit width Powered by Mozilla PDF.js v3.11

Document Information

Document Title Sound-Based Bird Localization Using Lightweight Convolutional Neural Networks and Internet of Things
Author / Organization Chakir Bouarouguene, Moufida Maimour, Ouahab Kadri, Abderrazak Benyahia, Eric Rondeau
Subject & Discipline Multidisciplinary Academic Open Research
Categories Science & Mathematics
Publication Year 2025
License / Rights Open Access / Creative Commons
Resource Verification Listed PDF (Verification Pending)
Original Host Website HAL Open Science (hal.science)
File Format application/pdf (Portable Document Format)
Language EN
Catalog ID DOC-256269
Original Source Link Visit Source Page on hal.science »