The BeeMate: Air quality monitoring through crowdsourced audiovisual data

The BeeMate module for collecting and analyzing audiovisual data related to air quality monitoring is presented. This module is implemented as part of a mobile citizen science application for user engagement and environmental behavioral change. The application allows volunteers to perform air quality measurements along with audiovisual data capturing in urban areas, using low-cost and built-in mobile device sensors. The BeeMate module implements a micro-service architecture, providing semantic analysis of the collected audiovisual data, and associating them with the detection of air-polluting sources. Several machine learning models are integrated into the backend of the application. An audio-driven model for the detection of air-polluting sources based on a 1D-CNN architecture is proposed and evaluated. Moreover, to address the problem of managing multi-source, multimodal, and heterogenous data, a lexicon of pollution-related terms is built.