AnimalHunt: AI-based animal sound recognition application

• Main Author: Lin, Yi Heng
• Other Authors: Owen Noel Newton Fernando
• Format: Final Year Project
• Language: English
• Published: Nanyang Technological University 2023
• Subjects: Engineering::Computer science and engineering::Computing methodologies::Artificial intelligence
• Online Access: https://hdl.handle.net/10356/165942
• Institution: Nanyang Technological University

This paper describes the development of an A.I. application for animal sound classification using pre-trained and custom-trained machine-learning models deployed on mobile devices. The research aims to address the challenges of traditional animal acoustic sound signal analysis, which is computationally intensive, requires a strong network connection, and is challenging to implement on low-cost microcontroller-based systems. By using Yet Another Mobile Network (YAMNet), a pre-trained model, and a custom-trained model, animal sounds and noises can be identified in real time, and the animal making the sound can be determined. The accuracy of the predictions is evaluated using a mobile device's trained model against test datasets in three different modes. Although the animal scope is currently limited to birds found in Singapore due to dataset constraints, the system can be expanded to other animals and species as long as sufficient datasets are available, making it a promising solution for continuous real-time biodiversity monitoring.