Automatic Baby Cry Diary

Crying is an infant’s earliest and most effective mode of communication. This communication process can be disrupted if cry characteristics or acoustic attributes associated with infant crying are abnormal. The standard of measurement in the study of infant cry abnormalities has been a written cry diary. Cry diaries produced in the home by parents produce inconsistent and unreliable data. In this work, baby-cry was recorded over a 24-hour period in the natural home environment and digitized for computer-based analysis. The various sounds that have comparable energy or overlap baby-cry both in time and frequency were included in the recordings. Our goal was to identify all baby cry time segments in order to automate the generation of a baby-cry diary. Our Automatic Baby Cry Diary model consists of a 12-D Feature Vector and a Neural Network for classification. Based on our results, we believe our Automatic Baby Cry Diary shows strong promise in developing practical baby-cry analysis tools to aid physicians in diagnosing infant diseases and conditions such as Colic.