Decoding and spatial mapping of acoustic noise in the neonatal intensive care unit

The neonatal intensive care unit (NICU) is a critical care setting where premature infants face continuous exposure to elevated noise levels, often exceeding international safety guidelines. While the risks of excessive acoustic exposure are well established, strategies for real-time noise monitoring and mitigation in operational NICUs remain underexplored. In this study, we propose an exploratory framework that integrates spatially distributed sound sensors, acoustic heatmap visualization, and machine learning-based classification to analyze and categorize noise events in a high-density NICU setting. The analysis identified persistent high-noise zones near incubators and entryways, with staff movement and alarm-related activities causing significant sound level spikes-particularly during the noon shift. Additionally, a random forest classifier achieved 85.5% accuracy in distinguishing clinical activity patterns based on environmental acoustic data. While not intended for urgent alerting, this framework demonstrates the potential of using ambient sound profiles for non-critical event recognition and environmental monitoring in the NICU.