Analysis of Poultry Vocalizations for Age Estimation using Cepstral Coefficients and C4.5 Classification

摘要

Poultry scientists and researchers have been studying the possibility of identifying bird stress from the vocal patterns of broiler flocks. Being able to identify stress or sickness in a flock can lead to early interventions. However, as the flock ages, the vocal properties change. This paper presents the preliminary results of an age classification technique using the C4.5 decision three algorithm using Mel-frequency cepstral coefficients (MFCCs), audio frequencies commonly used from speech processing. Sound data was collected continuously over a 61-day period. The flock was composed of 400 Cobb500TM broilers raised under research conditions mimicking commercial operations. Results of the data processing and the C4.5 classification indicate that age can be determined within one day of error with 70% frequency. This process of determining age of a flock is intended to be the first step in a larger comprehensive system for identifying the condition of the flock and assessing animal welfare.