A tristimulus-formant model for automatic recognition of call types of laying hens

An essential objective of Precision Livestock Farming (PLF) is to use sensors that monitor bio-responses contain important information on the health, well-being and productivity farmed animals. In literature, vocalisations animals have been shown can enable farmers improve their animal husbandry practices. this study, we focus vocalisation birds specifically develop a sound recognition technique for continuous automatic assessment laying hen vocalisations. This study introduces novel feature called “tristimulus-formant” call types hens (i.e., types). Tristimulus considered be timbre equivalent colour attributes vision. measures mixture harmonics in given sound, which grouped into 3 sections according relative weights signal. Experiments were designed calls from 11 Hy-Line brown recorded cage-free setting (4303 labelled 168 h recordings). Then, processing techniques used extract features each type classify using LabVIEW® software. For extraction, focused extracting Mel frequency cepstral coefficients (MFCCs) tristimulus-formant (TF) features. two different classifiers, backpropagation neural network (BPNN) Gaussian model (GMM), applied recognise types. Finally, comparative trials test models. The results show MFCCs-12+BPNN (12 variables) had highest average accuracy 94.9 ± 1.6% but training time (3201 119 ms). At same time, MFCCs-3+TF+BPNN fewer dimensionalities (6 required less (2633 54 ms) than could well without compromising (91.4 1.4%). Additionally, BPNN classifier was better GMM recognising hens’ calls. chicken sounds effectively at low computational cost, giving it considerable potential large data analysis online monitoring systems.