Feature selection for improved classification

Author

Shaudys, Fred E. ; Leen, Todd K.

Author_Institution

Dept. of Comput. Sci. & Eng., Oregon Graduate Inst. of Sci. & Technol., Beaverton, OR, USA

Volume

4

fYear

1992

fDate

7-11 Jun 1992

Firstpage

697

Abstract

The authors apply the feature-selection technique of K. Fukunaga and W. Koontz (1970), an extension of the Karhunen-Loeve transformation, to spoken letter recognition. Feedforward networks trained for letter-pair discrimination with the new features showed up to 37% reduction in classifier error rate relative to networks trained with spectral coefficients. This performance increase was accompanied by a 91% reduction in feature dimension. For three-letter discrimination, the new features performed comparably to spectral coefficients, with a 90% reduction in feature dimension

Keywords

feedforward neural nets; pattern recognition; speech recognition; Karhunen-Loeve transformation; classification; classifier error rate; feature-selection technique; feedforward neural networks; letter-pair discrimination; spectral coefficients; spoken letter recognition; three-letter discrimination; Acoustic emission; Computer science; Data mining; Drives; Eigenvalues and eigenfunctions; Error analysis; Neural networks; Principal component analysis; Speech analysis; Training data;

fLanguage

English

Publisher

ieee

Conference_Titel

Neural Networks, 1992. IJCNN., International Joint Conference on

Conference_Location

Baltimore, MD

Print_ISBN

0-7803-0559-0

Type

conf

DOI

10.1109/IJCNN.1992.227237

Filename

227237