Presentation 1996/5/16
A Descrimination Method of Similar Characters Using Compound Mahalanobis Descrimination Function
Masato SUZUKI, Shin'ichiro OMACHI, Nei KATO, Hirotomo ASO, Yoshiaki Nemoto,
PDF Download Page PDF download Page Link
Abstract(in Japanese) (See Japanese page)
Abstract(in English) On handprinted chinese character recognition, one of the important problems is that some characters are confused with the ones having similar structure In order to distinguish from the similar characters, it is necessary to extract the characteristic parts from their feature vectors, and utilize them effectively. In this paper, we propose a new discrimination function, CMDF-Compound Mahalanobis Discrimination Function. Main feature of CMDF is the method of extracting the characteristic parts of similar characters using eigenvectors with small eigenvalues. In recognition experiment with ETL9B, most similar characters were recognized correctly by CMDF, and a very high recognition rate of 99.31% was obtained.
Keyword(in Japanese) (See Japanese page)
Keyword(in English) Handprinted Character Recognition / Similar Character / ETL9B / CMDF
Paper # PRMU96-4
Date of Issue

Conference Information
Committee PRMU
Conference Date 1996/5/16(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) A Descrimination Method of Similar Characters Using Compound Mahalanobis Descrimination Function
Sub Title (in English)
Keyword(1) Handprinted Character Recognition
Keyword(2) Similar Character
Keyword(3) ETL9B
Keyword(4) CMDF
1st Author's Name Masato SUZUKI
1st Author's Affiliation Department of Computer Science, Tokyo National College of Technology()
2nd Author's Name Shin'ichiro OMACHI
2nd Author's Affiliation Graduate School of Engneering, Tohoku University
3rd Author's Name Nei KATO
3rd Author's Affiliation Graduate School of Information Sciences, Tohoku University
4th Author's Name Hirotomo ASO
4th Author's Affiliation Department of Communication Engneering, Tohoku Univensity
5th Author's Name Yoshiaki Nemoto
5th Author's Affiliation Graduate School of Information Sciences, Tohoku University
Date 1996/5/16
Paper # PRMU96-4
Volume (vol) vol.96
Number (no) 40
Page pp.pp.-
#Pages 8
Date of Issue