Identifying Construction Sounds: Smart sensors to understand noise nuisance

How to better identify sources and reduce sound pollution in urban areas? This pilot uses smart sensors to monitor construction noise in Amsterdam.

Identifying Construction Sounds

Noise pollution is a growing concern in Amsterdam, with construction sounds being one of its leading contributors. It can lead to a range of health issues, including sleep disturbance, stress, and fatigue. Responsible Sensing Lab (RSL) in collaboration with the Emission-Free Mobility Team from the City of Amsterdam's Department of Space and Sustainability (R&D) (formally: Uitstootvrije Mobiliteit van Ruimte en Duurzaamheid) have developed a joint research proposal to monitor and better understand construction-related noise. By deploying smart technology, we aim to gain deeper insights into the nature of construction noise and lay the groundwork for responsible solutions and informed policymaking.

Too much noise in the city can lead to serious nuisance for residents and can also lead to health damage. The measurement program is aimed at gaining more insight into the construction noises that occur. A better insight can also lead to a reduction of nuisance in the long term, e.g. by using quieter techniques.”
— Carlo Schoonebeek, Senior policy officer R&D, Gemeente Amsterdam

Identifying Construction Sounds - Responsible Sensing Lab

Identifying Construction Sounds - Responsible Sensing Lab 2

Examples of construction sites in Amsterdam - Images by Girish Vaidya

Goal: Gaining deeper insights into construction noise

Traditional sound level meters measure decibel levels but offer limited context. Yet, noise nuisance often depends on the type and source of the sound. This pilot project, initiated by AMS Institute, seeks to move beyond decibel readings to explore the unique characteristics of construction noise.

Our goal is to develop a solution that not only registers construction sounds but also automatically identifies their specific sources. The project uses a novel sound sensor developed by TU Delft that detects sound, measures a range of acoustic features (like loudness and sharpness), and classifies them in real time.

This approach is particularly useful for understanding construction noise. After all, noise disturbance is not just about how loud a sound is—it’s also about what kind of sound it is (high or low pitch, steady or fluctuating, buzzing or pounding) and what causes it (such as pile driving or drilling).

This level of detail can provide valuable feedback to both the creators of the noise (e.g., construction companies) and the communities affected by it. For city departments like R&D, it may inform future noise regulation and planning.

Another key aim of the project is to further evaluate and refine the TU Delft sensor. A central question is whether construction noise can be effectively distinguished from other urban sounds, such as traffic.

Two common construction sounds:

Drilling (wav)

Piling (wav)

Project details

Each sensor node includes a consumer-grade microphone (ICS-43434) capable of measuring loudness and identifying sound types. All data processing takes place locally on the sensor, meaning no audio is recorded, stored, or transmitted. The sensor was introduced as open hardware during the Inter.noise 2024 conference in Nantes in August 2024, where it was presented to an international scientific audience.

Sensors will be deployed at four construction sites across Amsterdam Oost, Zuid, and West. Each site represents a different type of activity—new construction, demolition, tramway work, and renovation—ensuring a diverse set of construction sounds is captured. The deployment is scheduled for four weeks, starting in May 2025.

sound sensor developed by TU Delft that detects sound, measures a range of acoustic features (like loudness and sharpness), and classifies them in real time.

Sensors are placed on 4 locations in Amsterdam - Images by Girish Vaidya

Responsible data collection and analysis

Ethical considerations are central to this project. All acoustic metrics are processed directly on the sensor, with no raw audio being saved or transmitted. In addition, the sensor’s design—both hardware and software—will be fully open-source to promote transparency and collaboration across the research community.