Infant Weeping Calls Decoder using Statistical Feature Extraction and Gaussian Mixture Models

Communication is the most important process for a living organism as this is the medium for the information exchange. But for the infants, the only way of communication is through crying. In this way, the baby expresses own needs. There can be many reasons for a baby to cry like unfed, sleep, tired, or uneasiness. The skilled parents can be easily distinguished about their baby needs by its actions and crying. But for the novice parents it becomes too difficult. This research area shows that around 60% of parents become frustrated when they can't figure out on why their baby is crying. According to Pew Research “In 2016, moms spent about 25 hours a week on paid work, compared with nine hours in 1965”. It shows that humans are becoming busier and especially mothers, who play a very important role in personality development of a baby, are working 25 hours in a week. Therefore this research area is very important for the baby growth. In this paper the initial results are obtained by applying various methods that were successful in speech and language recognition to the task of baby cry recognition. This paper is proposed a method based on the Gaussian Mixture Model for the baby communication. The results are promising 81.27% accuracy on a dataset with 714 babies and 5 types of crying.