Using clustering methods to detect quality data in a smartphone-based crowd-sourced database for environmental noise assessment

Abstract

Environmental noise is a major source of annoyance with serious effects on health. Therefore, noise assessment is crucial to reduce these impacts. An alternative approach has been developed (i.e. noise measurement with smartphones) to overcome the limitations of classical assessment methods (e.g. simulation tools or noise observatories). In this way, the NoiseCapture application consists of measuring and sharing data, in order to produce community noise maps. Nevertheless, collected data may suffer from problems such as a lack of calibration, which lowers its quality. Quality control is therefore very important to enhance the data analysis and the relevance of the noise maps. Having trustworthy data as a reference can help in assessing the database, for example using machine-learning methods. WIth NoiseCapture, such data can be collected thanks to a NoiseCapture Party, an organized event, on limited space/time (i.e. a cluster of data). Because not all events are known by the people in charge of NoiseCapture, and since the corresponding data can be considered of better quality, so their detection is a relevant task to increase the trust database. In the present communication, a clustering methodology is then proposed to automatically detect data that could be produced in such events.