Identification of atypical sound events in urban area

This paper reports themedium-term results of urban noise monitoring activities, that are still in progress in the city of Milan. In order to acquire acoustic data that provide a true representation of the sound emitted by roads, a phase of data validation is required. Specifically, adverse weather conditions and acoustic irregularities have to be recognized. A noise data set (consisting in daily and weekly series) has been constructed with an extensive monitoring system composed by permanent and mobile terminals. The first statistical analysis is operated on hourly equivalent levels and allows to identify the potentially anomalous one hour periods and to correct the average daily and weekly noise levels related to the specific sound source (in our case, the urban road traffic). The second analysis is operated on sound spectra, comparing them with the average ones. This allows to characterize the nature of the event and assign a level to a specific type of abnormal event, which can be classified as inherent or extraordinary with respect to the investigated source. Working on a significant statistical basis allows to implement a methodology for automatic detection of extraordinary acoustic events in the noise urban monitoring