Title: Pattern recognition and separation of road noise sources by means of ACF, MFCC and probability density estimation
Authors: VALERO GONZALEZ XavierALÍAS PUJOL FrancescKEPHALOPOULOS StylianosPAVIOTTI Marco
Citation: EURONOISE 2009 Conference vol. 1 p. 1-9
Publisher: Institute of Acoustics
Publication Year: 2009
JRC N°: JRC57493
URI: http://www.euronoise2009.org.uk/
http://publications.jrc.ec.europa.eu/repository/handle/JRC57493
Type: Contributions to Conferences
Abstract: Noise source separation is a key issue in environmental noise assessment and of particular interest in the implementation of the European Environmental Noise Directive (END, 2002/49/EC), since according to the END the contribution of each single noise source to the overall noise level should be evaluated separately. Therefore, studies were performed in the context of an exploratory research project funded by the Joint Research Centre of the European Commission, with the main objective of investigating on different techniques that could automatically recognise and separate signals of different noise sources. These sources might be cars, trucks, scooters and background noise that commonly appear in real environments but are mixed. As a first step, an automatic recognition system has been fully developed under Matlab platform. The system is composed of two blocks. The first is a signal processing block, which parameterises the acoustic signals using several parameters extracted from the autocorrelation function, Mel energy coefficients or Mel Frequency Cepstral Coefficients. The second is a recognition block, which is based on the probability density estimation of those parameters. In order to check the system's performance, several tests were conducted using audio signals recorded in real environments. In those tests, autocorrelation function parameters showed the best results, with averaged recognition rates close to 80%.
JRC Institute:Institute for Health and Consumer Protection

Files in This Item:
There are no files associated with this item.


Items in repository are protected by copyright, with all rights reserved, unless otherwise indicated.