A comparative study of noise event identification using AI in unattended monitoring

This paper explores the use of multi-microphone devices and artificial intelligence (AI) for identifying noise events in unattended noise monitoring. The primary focus is to assess the reliability of a machine learning model initially trained on a dataset representing a particular soundscape. We evaluate the performance of this model when applied to diverse datasets collected from similar yet distinct soundscapes, encompassing various environmental conditions and noise profiles. Through comparative analysis, we determine the model's adaptability and potential limitations. The findings of this study offer insights into how well AI-based noise event identification models can work in different situations. This lays the groundwork for enhancing their applicability in diverse real-world settings and improving how well unattended noise monitoring systems function.