Classification of Typical Tursiops Aduncus Whistle Signals Using Convolutional Neural Networks

Dolphins, recognized as highly intelligent marine mammals, exhibit sophisticated communication and echolocation systems. Precise classification of dolphin whistles is pivotal for comprehending their communicative behaviors and monitoring their population dynamics, including size, structure, and distribution. This study presents the development of an extensive and high-quality dataset of dolphin whistle signals, sourced from the Chimelong Ocean Kingdom. This dataset includes unique whistle types that were previously unavailable to the research community. We investigate the application of Convolutional Neural Network (CNN) models to classify the whistle signals of the Indo-Pacific bottlenose dolphin (Tursiops aduncus). Multiple CNN architectures are employed to analyze and categorize these whistle signals. The performance of these models is evaluated using mean Average Precision (mAP), demonstrating that CNN-based methodologies can effectively distinguish between different dolphin whistle signals. This work provides valuable tools for marine biologists and researchers specializing in animal acoustics, enhancing the understanding of dolphin communication. It also contributes to the conservation and management efforts of dol-phin populations, offering significant insights into their behavior and ecological needs.