Decoding Infant Cry Using Audio Data and Machine Learning Approaches

Phursule, Kalyani Rajesh (2024) Decoding Infant Cry Using Audio Data and Machine Learning Approaches. Masters thesis, Dublin, National College of Ireland.

Preview	PDF (Master of Science) Download (1MB) | Preview
Preview	PDF (Configuration Manual) Download (2MB) | Preview

Abstract

This paper showcases a robust machine learning and deep learning based approach to classify the different infant cries. The significance of accurately determining why an infant is crying is from parents and caregivers perspective provides a motivation for this research. To accomplish this, several models such as Support Vector Machine (SVM), K-Nearest Neighbors (KNN), Random Forest, AdaBoost, and XGBoost, and deep learning models like, Long Short-Term Memory (LSTM), Recurrent Neural Network (RNN), Convolutional Neural Network (CNN), and ensemble models to improve the final model performance. The Synthetic Minority Over-sampling Technique (SMOTE) was implemented to handle imbalance in the dataset. Random Forest and XGBoost models returned highest accuracy, with Random Forest with SMOTE outperformed all the models by achieving 99.6% accuracy and 99% F1 score. Ensemble models also performed well but slightly less accurate than separate models trained with SMOTE. These competitive results highlight the importance of addressing data imbalance for better performance of a model, specifically for minority classes. Although the research is able to train machine learning models that could classify infant cries with significant accuracy, there still are challenges such as limitations of computational requirements of the models and real time audio processing. The future work involves transformer models, registration of pathological cry types, and optimization for low-resource environments targeting real-time applications. This research provides a fair basis for the further improvement of smart baby monitoring systems in adapting automated infant care.

Item Type:	Thesis (Masters)
Supervisors:	Name Email Rustam, Furqan UNSPECIFIED
Uncontrolled Keywords:	Baby cry classification; Machine learning; Ensemble learning; Deep learning; MFCC; Audio features extraction; SMOTE; Data Imbalance
Subjects:	Q Science > QA Mathematics > Electronic computers. Computer science T Technology > T Technology (General) > Information Technology > Electronic computers. Computer science H Social Sciences > HQ The family. Marriage. Woman > Children Q Science > Q Science (General) > Self-organizing systems. Conscious automata > Machine learning
Divisions:	School of Computing > Master of Science in Data Analytics
Depositing User:	Ciara O'Brien
Date Deposited:	04 Sep 2025 10:10
Last Modified:	04 Sep 2025 10:10
URI:	https://norma.ncirl.ie/id/eprint/8773

Actions (login required)

View Item