Presentation 2009-12-17
Accelerated Character Recognition Using a Regression-Tree Molding
Takahiro OTA, Toshikazu WADA,
PDF Download Page PDF download Page Link
Abstract(in Japanese) (See Japanese page)
Abstract(in English) High-speed and accurate character recognition is required for monitoring the condition of Ink Jet Printer (IJP) for industrial use, which is mainly used in production lines. However, there is no system that meets these requirements in commercially available character recognition systems. In this paper, we first propose, "modified compound similarity method", which is a high-accuracy but slow classifier. Next, we propose its acceleration method by learning its input-output relationship using linear regression trees and replacing the classifier by the learnt regression tree. We call this method is "Molding". For the molding a large amount of learning data is necessary. Then, we generated training data that imitates the behavior of IJP's feature variation. In the experiment, we confirmed that 500-1000 times speed up is achieved without dropping the recognition rate.
Keyword(in Japanese) (See Japanese page)
Keyword(in English) Subspace method / Compound similarity method / PaLM-tree
Paper # PRMU2009-134
Date of Issue

Conference Information
Committee PRMU
Conference Date 2009/12/10(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) Accelerated Character Recognition Using a Regression-Tree Molding
Sub Title (in English)
Keyword(1) Subspace method
Keyword(2) Compound similarity method
Keyword(3) PaLM-tree
1st Author's Name Takahiro OTA
1st Author's Affiliation Wakayama University:Kishugikenkogyo Co., Ltd()
2nd Author's Name Toshikazu WADA
2nd Author's Affiliation Wakayama University
Date 2009-12-17
Paper # PRMU2009-134
Volume (vol) vol.109
Number (no) 344
Page pp.pp.-
#Pages 6
Date of Issue