Assessment of impact of the learning sample volume on probabilities of stationary process recognition in the spectral domain

A. V. Yegorov; V. S. Parshin

doi:10.3103/S073527270502007X

Authors

A. V. Yegorov Ryazan State Radio-Engineering Academy, Russian Federation
V. S. Parshin Ryazan State Radio-Engineering Academy, Russian Federation

DOI:

https://doi.org/10.3103/S073527270502007X

Abstract

Analytical relationships are derived permitting to determine the number of realizations necessary for establishment of weight coefficients of the parametric decision rule, which ensure a preset probability of recognition. The results of numerical simulation are presented making it possible to estimate the recognition reliability when determining the parameters of the decision rule based on a single learning realization.

References

FOMIN, Y.A.; TARLOVSKII, G.R. Statistical Theory of Pattern Recognition [in Russian]. Moscow: Radio i Svyaz’, 1986.

MARPLE Jr., S.L. Digital Spectral Analysis with Applications. Prentice-Hall, 1987.

JENKINS, G.M.; WATTS, D.G. Spectral Analysis and Its Applications, Vol. 1. Holden-Day, 1969.

LEVIN, B.R. Theoretical Foundations of Statistical Radio Engineering [in Russian]. Moscow: Radio i Svyaz’, 1989.

REPIN, V.G.; TARTAKJOVSKII, G.P. Statistical Synthesis at a Priori Indeterminacy and Adaptation of Information Systems [in Russian]. Moscow: Sov. Radio, 1977.

GRADSHTEIN, I.S.; RYZHIK, I.M. Table of Integrals, Sums, Series, and Products [in Russian]. Moscow: Fizmatgiz, 1971.