ANN-based model to identify noticed sound events. A tool for inclusion in action plans against environmental noise in urban environments

Over recent years, acoustic pollution has become a common problem for urban environments and its treatment one of the new challenges in environmental policy. In this framework, the enactment of the European Directive on Environmental Noise Management (Directive 2002/49/EC) set forth the need to undertake strategic action plans to improve the sound conditions of cities. According to this point, based on the noise mapping results, competent authorities are required to develop action plans to reduce noise where necessary and maintain environmental noise quality where it is good. In developing plans of action, the competent authorities are faced with the problem of determining which specific measures can be addressed from a technical and economical standpoint. An economically and technically feasible action to undertake to improve problems identified in terms of noise pollution in urban agglomerations could be the identification and elimination of noticed sound events present in these environments. Noticed sound events are loaded with large amounts of sound energy, creating a significant increase above the sound climate where they appear and, therefore, behaving within the sound space as highly disruptive elements. For this reason, in this paper a tool based on artificial neural networks for the automatic identification of noticed sound events in monitored records of sound pressure level has been developed. To build the ANN-based model several types, structures and configurations of neural networks have been analyzed. The final goal is that the developed tool provides the information required for the implementation of a noticed sound event management plan