Título:	Deep learning and acoustic parameter analysis for identifying cattle vocalizations under confinement and handling conditions.
Autoria:	PAIVA, B. T. FERREIRA. A. P. L. PEREZ, N. B.
Afiliação:	BRYAN TEIXEIRA PAIVA, UNIVERSIDADE FEDERAL DO PAMPA; ANA PAULA LÜDTKE FERREIRA, UNIVERSIDADE FEDERAL DO PAMPA; NAYLOR BASTIANI PEREZ, CPPSUL.
Ano de publicação:	2025
Referência:	Sadio Electronic Journal of Informatics and Operations Research, v. 24, n. 2, p. 33-56, 2025.
Conteúdo:	Livestock production systems play a fundamental role in the Brazilian economy, serving as one of the country’s primary sources of income. During various stages of production – including veterinary procedures, weighing, and transportation – animals are subjected to different levels of handling, with the potential to cause stress. Animal stress significantly impacts meat properties, including reducing its quality to the point of making it unfit for human consumption. This paper analyzed cattle vocalizations under two psychologically distinct conditions: confinement and handling. The main objective is to identify stress-inducing situations using two approaches: analyzing the vocalizations’ acoustic parameters and applying them to train a deep learning network to learn the sound patterns emitted by the animals in each situation. In the acoustic analysis, a statistical study was conducted on the parameters of fundamental frequency (F0), spectral formants (F1–F4), jitter, shimmer, harmony, and intensity. For the deep learning study, three convolutional neural network architectures were implemented, using Mel Frequency Cepstral Coefficients (MFCC) for acoustic feature extraction. The results of the acoustic analysis revealed significant differences (p<0.001) between the parameters of stressed and non-stressed vocalizations for most parameters. Meanwhile, the neural network results show that the basic, intermediate, and robust architectures achieved F1-scores of 96.97%, 97.90%, and 98.74%, respectively.
Thesagro:	Bem-Estar Bovinocultura Bovino Agricultura de Precisão
ISSN:	1514-6774
Tipo do material:	Artigo de periódico
Acesso:	openAccess
Aparece nas coleções:	Artigo em periódico indexado (CPPSUL)