A Low-Cost IoT-Based System for ContinuousUrban Noise Classification & Monitoring
2026 (English)Independent thesis Advanced level (degree of Master (Two Years)), 20 credits / 30 HE credits
Student thesis
Abstract [en]
Environmental noise pollution is a growing public health concern in European cities, yet continuous monitoring remains limited by the high cost, power demands, and infrastructural complexity of traditional sound level meters. This thesis presents the design, implementation, and evaluation of a low‑cost IoT-based system for continuous urban noise classification and monitoring in Malmö, Sweden. The proposed architecture integrates a MEMS microphone, embedded machine learning, and LoRaWAN communication to capture environmental audio, estimate A-weighted sound pressure levels, classify dominant noise sources, and transmit compact metadata suitable for autonomous outdoor deployment.
A custom 3D‑printed outdoor enclosure was developed and evaluated for weather resistance and acoustic impact. The system was assessed through experiments covering microphone calibration, enclosure-induced amplification, classification performance across multiple dataset iterations, transmission reliability in real urban environments, and operational feasibility under low-power constraints. The embedded classifier was trained using a combination of public datasets and MEMS-recorded audio, ultimately achieving high accuracy when aligned with the acoustic characteristics of the deployed hardware.
Results demonstrate that low-cost sensing hardware can provide stable dB(A) measurements after calibration and that embedded classification, combined with transmission-efficient payload design, enables meaningful factor-based interpretation of urban noise. LoRaWAN testing showed that lightweight metadata transmission is feasible with acceptable packet success rates for non-real-time monitoring. Although challenges remain regarding long-term power supply, enclosure acoustics, and domain-specific dataset refinement, the findings indicate that affordable IoT systems can complement traditional noise mapping frameworks and support scalable, continuous environmental noise monitoring in medium-sized European cities.
Place, publisher, year, edition, pages
2026. , p. 69
Keywords [en]
Internet of Things (IoT), Urban Noise Monitoring, Environmental Sound Classification, Artificial Intelligence, LoRaWAN
National Category
Computer and Information Sciences
Identifiers
URN: urn:nbn:se:mau:diva-88644OAI: oai:DiVA.org:mau-88644DiVA, id: diva2:2110176
Educational program
TS IoT- Internet of Things
Supervisors
Examiners
2026-10-152026-10-132026-10-15Bibliographically approved