Dynamic Time Warping for classifying cattle behaviors and reducing acceleration data size

Abstract

This paper proposes a method for classifying the cattle behaviors. An embedded accelerometer system has attached to the cow’s neck. Dynamic Time Warping (DTW) is applied to measure the similarity between acceleration data corresponding to the cow movements and the templates collected from the acceleration data corresponding to the cow behaviors. The results of these processes are the sets of accumulated distances whose minimum value is used to select a behavioral model. Two cows used in the experiment, the accuracy of classification was measured. The results show that the accuracy of the proposed system is more than 90 percentages for all behavioral models. Moreover, the three-axis acceleration data combined before sending through the wireless network to the computer base results in the power consumption of wireless network reduced.