A Comparative Analysis: Enhancing Baby Cry Detection with Hybrid Deep Learning Techniques

Reliability in identifying baby cries in various auditory contexts has significant consequences for infant welfare and caretakers. To identify and highlight spectral features in audio data, this study uses Mel Frequency Cepstral Coefficients (MFCCs), with a focus on distinguishing baby cries from other noises. This study's main goal is to improve sound classification accuracy, especially when it comes to separating newborn cries from background noise. The primary result of this work is the combining of Convolutional Neural Networks (CNN) and Long Short-Term Memory (LSTM) networks to achieve an astounding 99% accuracy in sound categorization. This combination performs substantially better than the individual CNN and LSTM models. This result indicates how well the suggested method works to recognize infant cries, and it has great potential for use in a variety of childcare and surveillance settings.