Audio classification for radio broadcast indexing: feature normalization and multiple classifiers decision

Authors:
Christine Sénac;Eliathamby Ambikairajh
Affiliations:
Institut de Recherche en Informatique de Toulouse, UMR 5505 CNRS INP UPS, Toulouse Cedex 04, France;School of Electrical Engineering and Telecommunications, University of New South Wales, Sydney, Australia
Venue:
PCM'04 Proceedings of the 5th Pacific Rim Conference on Advances in Multimedia Information Processing - Volume Part II
Year:
2004

Citing 4
Cited 0

Transcriber: Development and use of a tool for assisting speech corpora production

Speech Communication - Special issue on speech annotation and corpus tools
Information fusion in biometrics

Pattern Recognition Letters - Special issue: Audio- and video-based biometric person authentication (AVBPA 2001)
Audio Segmentation and Classification based on a Selective Analysis Scheme

MMM '04 Proceedings of the 10th International Multimedia Modelling Conference
Speech/music discrimination for multimedia applications

ICASSP '00 Proceedings of the Acoustics, Speech, and Signal Processing, 2000. on IEEE International Conference - Volume 04

Quantified Score

Hi-index	0.00

Visualization

Abstract

This paper presents a system that detects the two basic components (speech and music) in the context of radio broadcast indexing. The originality of the approach covers three different points: a differentiated modelling based on Gaussian Mixture Model (GMM), which permits the extraction of speech and music components separately, the normalization of commonly used features and the efficient fusion of classifiers for speech classification which provides a substantial improvement in the presence of strong background music: accuracy of the indexing system goes from [69.2%,94.2%] for the best classifier to [90.25%,98.56%] for the fusion. Evaluation was performed on 12 hours of radio broadcast recorded under various noise conditions, channels and containing diverse speech and music mixtures.