Sinnott R. Predicting and Avoiding Dog Barking Behaviour through Deep Learning. ACSW '24: Proceedings of the 2024 Australasian Computer Science Week. 2024; Pages 26-35. https://doi.org/10.1145/3641142.3641176
Summary: This study addresses the challenge of excessive dog barking by using deep learning to predict environmental sounds that trigger barking. The system captures Mel-Frequency Cepstral Coefficients and Mel Spectrogram features from urban audio datasets, using convolutional neural networks to classify 50 common sound triggers. With an accuracy of 87.6%, the model helps predict sounds likely to provoke barking. The goal is to replace negative reinforcement, such as reprimands, with automated positive reinforcement, rewarding dogs when they remain quiet, thus modifying behavior even when owners are absent.
Commentaires