Use of vocalisation to identify sex, age, and distress in pig production

• Published in: Biosystems engineering 2018-09, Vol.173, p.57-63
• Main Authors: Cordeiro, Alexandra F.da S., Nääs, Irenilza de A., da Silva Leitão, Felipe, de Almeida, Andréia C.M., de Moura, Daniella Jorge
• Format: Article
• Language: English
• Subjects: Acoustic signals; Machine learning; Pig production; Pig welfare
• ISSN: 1537-5110; 1537-5129
• DOI: 10.1016/j.biosystemseng.2018.03.007

To assess animal welfare at a pig production farm is a time-consuming task. The present study aimed to investigate the differences in pig vocalisation as a function of the sex, age, and distress conditions, and to propose a way of identifying distressful situations. The individual vocalisations of 40 pigs were recorded (20 male and 20 female) during exposure to different distress in the farrowing, nursery, growth, and finishing phases. Vocalisation pitch differed between males (194.5 Hz) and females (218.2 Hz). Pig vocalisation was also different according to age, especially for the attributes of maximum and minimum amplitudes, and the frequency of formant 2. Diverse distress situations also were identified by various acoustic attributes. A decision-tree for classifying the distress condition for pigs was built (with an accuracy of 81.92%) using the machine-learning technique. Results indicate the possibility of estimating pig welfare by recording the vocalisation. The algorithm is also promising to identify pig sex and age. •Formant 1 and pitch were most reliable acoustic parameters to estimate pig age.•Pigs change their vocal response as they grow.•Pitch was the most dependable variable for identifying if the pig was male or female.•Pigs change the vocalisation when exposed to a different distress situation.•A decision-tree to identify pig distress was built using machine learning technique.