Automatic detection of suckling events in lamb through accelerometer data classification

Abstract

We report on an experimental study aimed at establishing a framework for automated detection of suckling episodes in lamb. Suckling turns out to be an important element of the animal’s behavior, because it occurs early in its development cycle and is directly linked to the fundamental predictors of its success. Our objective was to build an inexpensive, unobtrusive, maintenance-free, and energy-efficient device easily attachable to the lamb that would reliably detect suckling episodes and report them wirelessly to a data collection point. We demonstrate that suckling is characterized by a rather simple and distinguished acceleration signature which makes it possible to detect the event with relatively simple techniques easily implementable within low-end microcontrollers. We propose an algorithm to this end and assess its performance on acceleration data obtained from animals in a farm environment. Our algorithm has been able to detect 95% of all (actual) suckling episodes with less that 10% false indications.