NoisySoundBD: A Geotagged Noise Audio Dataset from Dhaka, Bangladesh for Noise-Source Classification

Published: 24 September 2026| Version 1 | DOI: 10.17632/y35mpdn2jt.1
Contributors:
,
,
,
,
,

Description

NoisySoundBD is a geotagged and annotated urban noise dataset designed for noise-source classification. It contains 777 short environmental-audio clips collected from real-world urban settings, each paired with structured metadata such as location, timestamp, decibel statistics, recording context, severity, and human-reviewed labels. The dataset is organized into three broad classes: construction_industry, misc, and traffic. It can support research in urban noise monitoring, environmental audio classification, geospatial noise analysis, and low-resource audio machine learning. For more details and benchmarking performance, please read our research paper (link is provided).

Files

Steps to reproduce

Please cite our research paper (link is provided) if you use this dataset.

Institutions

Categories

Noise in the Environment, Urban Planning, Machine Learning, Acoustic Signal Processing, Deep Learning

Licence