Accelerometer based activity monitoring system for behavioural analysis of free-roaming animals

Abstract

In this paper, an activity monitoring system is described based on a custom wireless 3-axis accelerometer sensor. The system is suitable for behavioral analysis of larger free-roaming animals. The wireless sensor is mounted on a collar and transmits the accelerometric data to the PC logging application. The received data can be recorded fully or partially by segmenting the events upon derivative threshold crossing or manual triggering. Furthermore, a segmentation method is proposed based on periodicity extraction from the selected axis data. The segmented data chunks can be introduced to machine learning algorithms (Artificial Neural Networks, Hidden Markov Models or Support Vector Machines) for further classification.