Analysis of cycles in urban noise based on long term data

Long term monitoring data can be utilised for detecting patterns in variations of noise levels. Approaches based on the Fourier transform technique are too sensitive to the quality of data, which may be affected by rain, periods with high wind speeds and extraneous noises. Accurate detection of long cycles in environmental noise using conventional methods requires a long, uninterrupted period of data acquisition. This can rarely be carried out in practical situations. Wavelet analysis is widely used as a denoising technique which allows for the reliable detection of periodic components in a complex signal. This technique is also relevant for analysis of non- stationary processes. This paper details a method which allows the extraction of major periodic components utilizing a set of data which is not rectified in accordance with typical procedures, e.g. contains all sound pressure level estimates independently of weather conditions. Correspondingly long time histories can be post- processed without interrupting the gathering of data. This method is explored on sets of long term data acquired from multiple noise monitoring stations in urban areas. Periods and relative magnitudes of intraday and longer cycles are analysed. It is shown that all major intraday cycles are similar for all of the monitoring stations and have similar relative magnitudes. There is a greater diversity in longer cycles of urban noise. The suggested technique of data analysis and information on the noise cycles can be used for analysing cycles in environmental noise, confirming contribution from a range of sources or designing generic patterns of noise level variations in particular areas.