Development of Assistive Technology for Pain Detection in Neonates

Assessing pain in neonates is a critical challenge due to their inability to communicate verbally. Traditional pain assessment methods rely on observational scoring, which is subjective and inconsistent. This study proposes a machine learning-based framework for automated neonatal pain detection using acoustic features extracted from infant cries. The system employs (Mel-frequency cepstral coefficients) MFCC, chroma features, and spectral contrasts for feature extraction, followed by classification using XGBoost, Random Forest, and Support Vector Machines (SVM). To address class imbalance, Synthetic Minority Oversampling (SMOTE) is applied. Experimental results indicate that SVM tuned achieves the highest accuracy of 98.9%, outperforming Random Forest (97.38%) and XGBoost (95.29%). The model also demonstrates superior precision (0.95) and recall (0.94), ensuring reliable pain classification. The proposed assistive technology provides a scalable and cost-effective solution for real-time neonatal pain assessment, aiding healthcare professionals in timely and objective medical interventions.