Handwritten numeral recognition using neural networks: improving the accuracy by discriminative training

Author

Liu, Cheng-Lin ; Nakagawa, Masaki

Author_Institution

Venture Bus. Lab., Tokyo Univ. of Agric. & Technol., Japan

fYear

1999

fDate

20-22 Sep 1999

Firstpage

257

Lastpage

260

Abstract

Artificial neural networks have been widely used in handwritten numeral recognition (HNR) with some success. This paper presents some new results of HNR using MLP (multilayer perceptron) and RBF (radial basis function) neural networks. By using discriminative training, which aims to minimize the empirical classification error on a training data set, the recognition accuracy is considerably improved. The performance of the RBF net is comparable to that of the MLP in terms of the forced recognition rate, and even better than the MLP in terms of the rejection-error tradeoff. The experiments were implemented on the CENPARMI database, and very high recognition rates have been obtained

Keywords

feature extraction; handwritten character recognition; learning (artificial intelligence); multilayer perceptrons; performance evaluation; radial basis function networks; CENPARMI database; discriminative training; empirical classification error minimization; forced recognition rate; handwritten numeral recognition; multilayer perceptron; performance; radial basis function neural network; recognition accuracy improvement; rejection-error tradeoff; training data set; Artificial neural networks; Computer science; Feature extraction; Filters; Handwriting recognition; Image sampling; Laboratories; Multi-layer neural network; Neural networks; Shape;

fLanguage

English

Publisher

ieee

Conference_Titel

Document Analysis and Recognition, 1999. ICDAR '99. Proceedings of the Fifth International Conference on

Conference_Location

Bangalore

Print_ISBN

0-7695-0318-7

Type

conf

DOI

10.1109/ICDAR.1999.791773

Filename

791773