The analysis of how the choice of membership functions influences the quality of recognition system

Author

Aliyeva, Nigar ; Ismayilov, Elviz

Author_Institution

Inst. of Inf. Technol., Baku, Azerbaijan

fYear

2012

fDate

2-4 July 2012

Firstpage

1

Lastpage

3

Abstract

Was analyzed effectiveness of the hand-printed recognition system using neuro-fuzzy network with membership functions from different classes. For pattern features was constructed smooth membership functions, and also known triangular, trapezoidal, Gaussian and bell membership functions. Analyzed results of numerical experiments of recognition system with membership functions from different classes. Revealed comparative advantage of smooth membership functions offered by authors.

Keywords

fuzzy neural nets; handwritten character recognition; Gaussian membership function; bell membership function; hand-printed recognition system; neuro-fuzzy network; smooth membership function; trapezoidal membership function; triangular membership function; Character recognition; Feature extraction; Fuzzy neural networks; Fuzzy sets; Neural networks; Text recognition; Training; feature extraction; hand-printed recognition system; neuro-fuzzy network; smooth membership functions;

fLanguage

English

Publisher

ieee

Conference_Titel

Innovations in Intelligent Systems and Applications (INISTA), 2012 International Symposium on

Conference_Location

Trabzon

Print_ISBN

978-1-4673-1446-6

Type

conf

DOI

10.1109/INISTA.2012.6247019

Filename

6247019