Audio classification based on machine learning: understanding animal behavior through sound

Date

2023-9-06

Author

Yaz, Furkan

Metadata

Show full item record

This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License.

Item Usage Stats

316
views

166
downloads

Cite This

Machine learning-based products that try to make our lives easier are increasing day by day. Thanks to the machine learning models running behind them, these products can be seen or heard and provide information about the context they are in. In this study, three machine learning methods that can hear and understand cat sounds were developed to serve this purpose. These models are Artificial Neural Network, Convolutional Neural Network, and CatBoost. A data library of two thousand sounds was created to understand 6 different cat behaviors. The process proceeded in three basic steps: pre-processing, feature extraction, and classification. One of the most widely used feature extraction algorithms, Mel-Frequency Cepstral Coefficients (MFCCs) has been preferred for Audio Feature Extraction. Accuracy was used as the evaluation metric. A minimum classification success rate of 95% was achieved in all models and the most successful model was determined as Convolutional Neural Network. The reason for this is considered as the presence of convolutional and pooling layers in the model architecture.

Subject Keywords

Audio classification, Artificial neural network, Convolutional neural network, CatBoost, Machine hearing

URI

https://hdl.handle.net/11511/105398

Collections

Graduate School of Natural and Applied Sciences, Thesis