Detecting cetaceans, such as dolphins and whales, using their acoustic signals (i.e., vocalizations) is an important tool for their monitoring. This thesis looks at how embedded machine-learning models can help identify cetacean sounds in real-time from signals coming from hydrophones (i.e., underwater microphones), to create a system that is both efficient and scalable for use in diverse marine environments, especially onboard monitoring ships. The acoustic signals will be processed to extract key features using signal processing techniques and will be used to train machine learning models. These models will then be designed to recognize cetaceans, even with background noise from ship engines or natural disturbances. The models will be optimized for embedded systems, from both the needed hardware and computational resources. The expected outcome of this research is to show that these machine learning models can reliably identify cetacean sounds, even in challenging conditions. By combining embedded systems with machine learning, this work offers a cost-effective and scalable solution for continuous monitoring of marine life. Ultimately, this research aims to contribute to marine conservation efforts by providing a practical way to detect and monitor cetaceans in their natural habitats.
Embedded Machine Learning Models for the Detection of Cetaceans via Acoustic Signals
MAHIN MOHAMMADALIZADEH, REZA
2024/2025
Abstract
Detecting cetaceans, such as dolphins and whales, using their acoustic signals (i.e., vocalizations) is an important tool for their monitoring. This thesis looks at how embedded machine-learning models can help identify cetacean sounds in real-time from signals coming from hydrophones (i.e., underwater microphones), to create a system that is both efficient and scalable for use in diverse marine environments, especially onboard monitoring ships. The acoustic signals will be processed to extract key features using signal processing techniques and will be used to train machine learning models. These models will then be designed to recognize cetaceans, even with background noise from ship engines or natural disturbances. The models will be optimized for embedded systems, from both the needed hardware and computational resources. The expected outcome of this research is to show that these machine learning models can reliably identify cetacean sounds, even in challenging conditions. By combining embedded systems with machine learning, this work offers a cost-effective and scalable solution for continuous monitoring of marine life. Ultimately, this research aims to contribute to marine conservation efforts by providing a practical way to detect and monitor cetaceans in their natural habitats.File | Dimensione | Formato | |
---|---|---|---|
Mahin Mohammadalizadeh_Reza.pdf
embargo fino al 17/03/2028
Dimensione
723.26 kB
Formato
Adobe PDF
|
723.26 kB | Adobe PDF |
The text of this website © Università degli studi di Padova. Full Text are published under a non-exclusive license. Metadata are under a CC0 License
https://hdl.handle.net/20.500.12608/82743