Behaviour 2019
The investigation of vocalization in free-ranging groups of horses (Equus caballus) by using machine learning
Sakiho Ochi, Satoshi Hirata. Wildlife Research Center, Kyoto, Kyoto, Japan

For a better understanding of animal society, many studies on animal vocalization have been conducted. In this study area, researchers sometimes encounter with nonlinear dependencies and unclear interactions coming from the large data set, and thus they may fail to conform to the classical statistical methods which result in limiting the analytical power on those data. Recently, by using machine learning (ML) for acoustic studies, expected following features have been emerging: 1) automatic extracting of specific vocalizations from large data set on acoustic recordings, 2) extracting unknown acoustic features from those data sets. In this research, we aimed to extract and investigate the features of three types of vocalizations (whinny, nicker, blow) in intraspecific and interspecific communications of horses (Equus caballus) by using the ML. We attached small microphones on each horse and recorded their vocalizations in four situations: 1) separation from a familiar individual, 2) demanding food to human, 3) aggression to another individual, 4) free ranging. In this presentation, we will report the preliminary results of the investigation of the features of their vocalizations.