Enriching noise environment analysis with spectrally sensitive event detectors: potential and examples from a citizen science project

Enriching datasets with various indicators adds value to sound environment classification and allows analysing the relation between exposure and noise annoyance and sleep disturbance in more depth. This is in line with the WHO guideline report acknowledging the lack of event based indicators to evaluate the health impact of environmental noise on people. Specific event detectors add value in various ways. Firstly, including spectral windows makes event detection less sensitive to short-term variability of high frequent natural sources. Secondly, a dynamic threshold can adjust to local backgrounds. The event detection then becomes less sensitive to local features related to the diurnal pattern, the distance to curb side and other. In addition, the detectors can be made sensitive to match the source properties (expected emergence and minimum duration of the event). In this manuscript, the potential of spectrally sensitive event detectors will be illustrated on real-life data.