Cow Identification and Recognition of Condition with Its Vocalization

Abstract

The objective of this research is automatization of livestock management using the animal voice which may be considered as the expression of animal demands. In this paper, vocalizations of seven cows were analyzed under hungry and weaning conditions to recognize individual cow and condition important in livestock management. It was assumed that the voice was produced by the auto-regressive process and the frequency characteristics of voice were expressed with the linear prediction coefficients. These coefficients were utilized in the discriminant analysis using the Mahalanobis' generalized distance. The rate of exact recognition ranged from 23.4% to 77.8% for recognition of individual, and the average rate was 59.8%. The average rate of exact recognition was 92.4% for recognition of condition.