家畜の鳴き声による個体識別

抄録

Animals communicate their thoughts by their vocalization as a human does by speech. Animal vocalization contains much important information, such as individual characteristics and physical conditions.
We propose livestock management with vocalization. In this research, cattle were identified with their vocalization. In generally, when the cattle are satisfied with their breeding condition, they do not cry. They cry under the limited situation such as hunger, estrus, sickness of calf, weaning. The vocalization of the cattle under the hungry condition was analyzed. Vocalizations recorded on a DAT-recorder were converted into digital time signal by an oscilloscope. The vocalization signals were modeled with AR-model by MEM (Maximum Entropy Method). Discriminant analysis was performed with Mahalanobis' generalized distance using the estimated AR-parameters. Over all rate of correct discrimination was 71.4%.