Presentation 2006-10-20
Feedback Character Recognition Using Feature Modification by Neural Network
Kiyofumi Kusakabe, Yoshimasa Kimura,
PDF Download Page PDF download Page Link
Abstract(in Japanese) (See Japanese page)
Abstract(in English) This paper proposes the method of feedback recognition system in character recognition. The pattern which recognition reliability is low will be transferred to discrimination part and is performed a minute modification so as to bring its feature close to the features of reference patterns, and is recognized again. Modification of the feature vector is done by using neural networks. On the experiments using 3,036 different characters including the JIS level-1 Kanji character, the correct recognition rate has improved by repeating feedback. Moreover, it was clarified that the effect of feedback had appeared strongly in categories whose initial recognition rate is low. The principle of feedback that enables to recognize reject pattern using neural networks is confirmed by these results.
Keyword(in Japanese) (See Japanese page)
Keyword(in English) Character Recognition / Feedback / Neural Networks / Displaced Feature / Learning
Paper # PRMU2006-109
Date of Issue

Conference Information
Committee PRMU
Conference Date 2006/10/13(1days)
Place (in Japanese) (See Japanese page)
Place (in English)
Topics (in Japanese) (See Japanese page)
Topics (in English)
Chair
Vice Chair
Secretary
Assistant

Paper Information
Registration To Pattern Recognition and Media Understanding (PRMU)
Language JPN
Title (in Japanese) (See Japanese page)
Sub Title (in Japanese) (See Japanese page)
Title (in English) Feedback Character Recognition Using Feature Modification by Neural Network
Sub Title (in English)
Keyword(1) Character Recognition
Keyword(2) Feedback
Keyword(3) Neural Networks
Keyword(4) Displaced Feature
Keyword(5) Learning
1st Author's Name Kiyofumi Kusakabe
1st Author's Affiliation Kochi University of Technology()
2nd Author's Name Yoshimasa Kimura
2nd Author's Affiliation Kochi University of Technology
Date 2006-10-20
Paper # PRMU2006-109
Volume (vol) vol.106
Number (no) 301
Page pp.pp.-
#Pages 6
Date of Issue