Automated Infant Pain Detection from Acoustic Features of Baby Cries

This study aims to develop an automated system for classifying pain and comfort states in infants through the analysis of their cries. The research presents a novel pipeline that encompasses quality checking, selection, and pre-processing of infant cry audio files. Salient acoustic features are extracted from the processed data to train machine learning classification algorithms, including Support Vector Machines (SVM) and a Convolutional Neural Network (CNN) model leveraging transfer learning techniques. The automated pipeline addresses challenges in data preparation and feature extraction, crucial steps in developing robust classification models for infant cries. By employing both traditional machine learning approaches (SVM) and deep learning methods (CNN), this study compares the efficacy of different classification strategies in discerning infant emotional states. This research is conducted as part of a collaborative project between the University of Padova and L'Inglesina Spa, with potential applications in enhancing infant care products and practices. The findings of this study contribute to the growing field of automated infant state recognition and have implications for both academic research and practical applications in pediatric care.