Behavior Classification of A Grazing Goat in the Argentine Monte Desert by Using Inertial Sensors

Abstract

The knowledge generated by animal behavior studies has been gaining importance due to it can be used to improve the efficiency of animal production systems. In recent years, sensor-based approaches for animal behavior classification has emerged as a promising alternative for analyzing animals grazing patterns. In the present article it is proposed the use of a classification system based on inertial sensors for identifying a goat’s grazing behavior in the Argentine Monte Desert. The data acquisition system is based on commercial off-the-self devices. It is used to create a reliable dataset for performing the animal behavior predictions. By fixing the system on the head of a goat it was possible to log its movements when it was grazing in a natural pasture. A preliminary version of the dataset is evaluated using a classical statistical learning algorithm. Results show that goat activities can be predicted with an average precision value above 85% and a recall of 84%.