A Bayesian classifier for baby's cry in pain and non-pain contexts

Since the birth, the babies have a communicative intention and the cry is the main way they use to express their needs and feelings to their caregivers. Previous works have demonstrated that applying signal processing techniques to analyze the sound of these cries, it's possible determinate which features carry information about the context that evoked the cry. Therefore, the present study investigates a classification method to allow the automatic recognition of babies' cry. From of 25 cries recorded in a pain context and 25 cries in a non-pain context, a Bayesian method was applies to project a two-context cry classifier. Preliminary results in a group of 50 cries sounds, separated in training and test groups in different folds, indicate a reasonable performance classification technique, around 75% of correct trials.