A method for reducing dimensionality of the space of attributes of recognizable objects

Authors

  • V. N. Bronnikov Donetsk Institute of Railway Transport, Ukraine
  • V. I. Poddubnyak Donetsk Institute of Railway Transport, Ukraine
  • V. Yu. Larin Donetsk Institute of Railway Transport, Ukraine

DOI:

https://doi.org/10.3103/S0735272705020093

Abstract

A new method is suggested for reducing the dimension of the space of pattern attributes, with simultaneous improvement of separability of patterns’ classes. The method is based on application of a variant of analog nonlinear decoding of coded continuous messages and permits to diminish dimensionality of the space of attributes by a factor of several hundred. The possibility of raising the signal-to-noise ratio at the analog decoder is confirmed by computer-aided simulation of analog decoding.

References

FUKUNAGA, K. Introduction to Statistical Pattern Recognition, 2nd ed. Academic Press, 1990.

BRONNIKOV, V.N. Radiotekhnika, n.6, p.64-67, 1980.

VINTSYUK, T.K. Analysis, Recognition, and Interpretation of Speech Signals [in Russian]. Kiev: Naukova Dumka, 1987.

BORODIN, L.F. Introduction into Theory of Noise-Stable Coding [in Russian]. Moscow: Sov. Radio, 1968.

TIKHONOV, V.I. Statistical Radio Engineering [in Russian]. Moscow: Radio i Svyaz’, 1982.

KOLESNIK, V.D.; MIRONCHIKOV, Y.T. Decoding of Cyclic Codes [in Russian]. Moscow: Svyaz’, 1969.

Downloads

Published

2005-02-09

Issue

Vol. 48 No. 2 (2005)

Section

Research Articles