Developing an Artificial Neural Network for Modeling and Prediction of Temporal Structure and Spectral Composition of Environmental Noise in Cities

Noise pollution in large cities is an ever-growing problem, due to several factors: the increase in demographic density, the increase in the number of per capita devices, appliances and vehicles capable of generating loud noise, and the fact that society is getting used to higher noise levels. One of the most important factors that help us to explain this fact is the road traffic, since as is generally established, road traffic is the most important and generalized sound source in the urban zones of the developed countries. Generally speaking, this one is also, with difference, the sound source that produces more disturbances and nuisances on the urban residents. However, road traffic is not the only noisy source in urban environments: other noisy sources relating to construction work, commercial activity, recreation, etc. have been found. At the same time, sound spaces where road traffic does not have a direct incidence and in which natural and social sounds predominate, e.g. green areas, can be observed (Torija et al., 2010a). The European Directive 2002/49/EC on the Assessment and Management of Environmental Noise aims to create a common framework for assessing exposure to environmental noise in all Member States. With the use of indicators and evaluation methods harmonized the results will be grouped into strategic maps. These maps are designed to comprehensively assess noise exposure in a given area, or for overall predictions in that area. In addition, they will be the basis for developing both action plans and strategies in the fight against noise (Directive 2002/49/EC). For the development of assessment and achievement of the objectives stated in the above mentioned directive, from the European Commission the methods used to predict different emission sources present in urban environments (industrial noise, road traffic, railway traffic and aircraft traffic) are recommended (Commission Recommendation 2003/613/EC). All these methods are based only on the obtaining of the A-weighted energy-equivalent sound pressure level (LAeq). Nevertheless, any physical characterization of a sound environment calls not only for consideration of the A-weighted sound pressure level (LAeq), but also requires description of the temporal structure and spectral composition of the sound (Berglund & Nilsson, 2001; Botteldooren et al., 2006). These factors bear great weight in the perception of noise (Viollon & Lavandier, 2000; Berglund & Nilsson, 2001