AI-Based Acoustic Source Recognition on the Cadence Platform
Julien PREUILH, Thierry Mazoyer, Charles-Etienne Lamort, Jean-Florent CROS, Houda BENYAICH · NOISE-CON proceedings · 2025
Cadence is a monitoring platform designed for acoustic and vibration sensors, commonly used in Smart Cities applications to detect anomalous events that may cause nuisances. This study focuses on the acoustic domain, utilizing Class I sound level meters to capture environmental noise. These sensors generate alerts based on predefined trigger conditions. The goal is to leverage the solution thanks to AI to analyze each alert and identify its source. This paper presents examples of acoustic monitoring and provides a brief overview of the AI model used for source recognition. A key application is construction site monitoring: site managers must ensure their activities do not generate excessive noise alerts due to threshold exceedances. When an alert is triggered, the AI system determines whether the noise originates from the construction site or an external source. Source recognition is crucial in acoustic monitoring to differentiate between various noise sources such as construction machinery, passing vehicles, or honking and other classes of possible urban sounds. This distinction enables more accurate noise assessments, allowing for better regulatory compliance and targeted mitigation strategies in urban environments. By integrating AI with Cadence, this approach enhances noise monitoring and improves decision-making in urban noise management.