Computational Method for Segmentation and Classification of Ingestive Sounds in Sheep

1 In this work we propose a novel method to analyze and recognize automatically 2 sound signals of chewing and biting. For the automatic segmentation and classi3 fication of acoustical ingestive behaviour of sheep the method use an appropriate 4 acoustic representation and statistical modelling based on hidden Markov models. 5 We analyzed 1813 seconds of chewing data from four sheep eating two different for6 ages typically found in grazing production systems, orchardgrass and alfalfa, each 7 at two sward heights. Because identification of species consumed when in mixed 8 swards is a key issue in grazing science, we tested the possibility to discriminate 9 species and sward height by using the proposed approach. Signals were correctly 10 classified by forage and sward height in 67% of the cases, whereas forage was cor11 rectly identified 84% of the time. The results showed an overall performance of 82% 12 for the recognition of chewing events. 13