Comparative analysis of CNN and TNN models in environmental noise source identification

This paper presents a focused examination of environmental noise, an issue of relevance due to its implications for human health and well-being, compliant with ISO 1996-2:2017 standards. Moving beyond the traditional methodologies that primarily rely on level exceedance, our study integrates advanced machine learning techniques to address the challenges in noise source identification. With the advent of IoT Noise Monitoring Terminals, the conventional manual auditory analysis methods have become less feasible. Our research explores the application of convolutional neural networks (CNNs), a standard in environmental noise analysis, and assesses the emerging utility of Transformer Neural Network (TNN) models in this domain. The aim is to conduct an objective comparison of these models, applying them to identical datasets to determine their effectiveness in identifying noise sources. Through this analysis, the study seeks to contribute to the field of environmental acoustics by offering insights into the comparative strengths and limitations of CNN and TNN models.