Improved recognition performance for orthogonal sources

Author

Grigis, Sébastien ; Holobar, Ales ; Zazula, Damjan

Author_Institution

Ecole Centrale de Nantes, France

Volume

2

fYear

2003

fDate

22-24 Sept. 2003

Firstpage

153

Abstract

This paper deals with the problem of recognition of multiple orthogonal sources buried in highly superimposed observations. The known blind source separation (BSS) approach is upgraded with a separation of sources using a classification procedure. Single source contributions are looked for in spatial time-frequency distribution (STFD) of the observed signals. The classification is based on STFD matrices which are grouped in the orthogonal and similar classes. The resulting separation algorithm outperforms other known approaches, as well in accuracy as by lower computational complexity.

Keywords

blind source separation; computational complexity; matrix algebra; pattern recognition; time-frequency analysis; STFD matrices; blind source separation; computational complexity; orthogonal sources; recognition performance; separation algorithm; single source contributions; spatial time-frequency distribution; superimposed observations; time-frequency distribution; Added delay; Blind source separation; Computational complexity; Convolution; Distributed computing; Filtering algorithms; Signal processing; Signal processing algorithms; Source separation; Time frequency analysis;

fLanguage

English

Publisher

ieee

Conference_Titel

EUROCON 2003. Computer as a Tool. The IEEE Region 8

Print_ISBN

0-7803-7763-X

Type

conf

DOI

10.1109/EURCON.2003.1248171

Filename

1248171