Development of vocal recording and analysis system for laying hens and cow based-on cloud-computing

Vocalizations of laying hens could be a useful method to analyze behavior. We developed the vocal recording system for monitoring and the artificial intelligence analysis system based on cloud-computing for laying hens in this study, with the aim of classifying laying hen and cow sounds in South Korea. We classified laying hens' sounds to eight classes and cow's sounds to seven classes. Total 1,533 classified records were acquired and used for development of classifier models. We proposed two type of convolutional neural network was proposed for modeling to analyze the vocal class of hens and cow. Classification model based on 2-D ConVnet was better performance with a satisfied accuracy which shows 81.02% of laying hens and 91.02% of cow for validation classifier performance.