Author: Arnaud Philibert, work done March - August 1999.
- Rapport.pdf (Pdf-file, 44 pages. First ten pages introducing TUT can be ignored.)
- Appendix.pdf (Appendix, 5 pages.)
Keywords: speech / music discrimination, signal classification, audio content analysis, speech processing
The study concerns the automatic classification of audio signals into two classes: speech and music. According to the signal processing practice, this requires extraction, evaluation and selection of features and then selection of a suitable classification method.
These two subproblems were studied and are reviewed. Emphasis of this study is laid on two points: on literature review and on algorithm development.
A literature review was conducted on music and audio classification and several related areas of interest. An appropriate decomposition of the problem and the selection of an approach is first considered. Then the state-of-the-art of the research is represented and discussed, and promising directions for further work are indicated.
The main and maybe the most interesting part of this study concerns the development of algorithms first to calculate the features used and then to conceptualize a classifier. The performance of the method is evaluated by applying it on a database created in the course of this project, based on various kind of music and different langages. All the algorithms were implemented and simulated in Matlab environment.