An embedded system for automatic classification of neonatal cry

An implementation of neonatal cry classification on an embedded platform is presented in this paper. It was perceived that diagnosis of the cause of human baby cry can be aided with an electronic device which can help parents and doctors comfort their baby. The present work gives detail of such a mobile device which runs on Android operating system and includes an Android application for neonatal cry classification. The proposed Android application features automatic detection of neonatal cry, classification of the cry in four classes (“Hunger”, “Pain”, “Wet Diaper” and “Others”), displaying the cause of cry and taking user feedback on each cry. The objective of the device has been tested with an accuracy of around 56% and a confusion matrix on generated test result has been discussed in the results section.