The June 18, 2012 edition of ACM TechNews
New Software Forecasts Noise Levels in a Street
University of Granada (Spain) (06/13/12)
University of Granada researchers are using neural networks to help predict and analyze urban noise. The Approximate Reasoning and Artificial Intelligence research group has developed software for determining noise frequency and the type of noise in a given area, and says it is more accurate than existing forecasting models that are based on traditional mathematical methods that use a specific set of data. “This is the first system to apply soft computing methods in urban noise assessment, and there is scarce literature available on this method,” says project participant Natalia Genaro Garcia. “While many noise forecasting models have been developed in different countries, none of them is accurate enough.” The system predicts urban noise levels using a dataset, such as street type, road conditions, average speed of the vehicles passing by, and road works, with a reliability of 95 percent. The researchers say the tool also will be helpful in performing urban noise mapping projects. The team is now working to limit the number of variables needed to produce an accurate forecast of noise at the street level.