Presentation 2012-12-13
High-Accuracy Clustering Using LDA for Fast Japanese Character Recognition
Takahiro Sasaki, Hideaki Goto,
PDF Download Page PDF download Page Link
Abstract(in Japanese) (See Japanese page)
Abstract(in English) Recognizing text in natural scene images is very important to develop various systems such as an assis-tant device for visually-impaired people. Since computational resources are limited on such mobile devices, a fast and precise Optical Character Recognition(OCR) algorithm is needed. In feature vector based OCR systems, Nearest Neighbor (NN) search is often used, and its speed improvement is quite important. In this paper, we develop an OCR scheme with tree-based clustering technique with LDA (Linear Discriminant Analysis), and the second-ordercandidate reduction by linear search in lower dimensional spaces. The experimental results show that our proposed method runs 33 times faster than the conventional linear search, at mere 0.3% accuracy drop.
Keyword(in Japanese) (See Japanese page)
Keyword(in English) Japanese character recognition / nearest neighbor search / real-time processing / classification tree / Linear Discriminant analysis
Paper # PRMU2012-73,HIP2012-62
Date of Issue

Conference Information
Committee PRMU
Conference Date 2012/12/6(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) High-Accuracy Clustering Using LDA for Fast Japanese Character Recognition
Sub Title (in English)
Keyword(1) Japanese character recognition
Keyword(2) nearest neighbor search
Keyword(3) real-time processing
Keyword(4) classification tree
Keyword(5) Linear Discriminant analysis
1st Author's Name Takahiro Sasaki
1st Author's Affiliation Tohoku University()
2nd Author's Name Hideaki Goto
2nd Author's Affiliation Tohoku University
Date 2012-12-13
Paper # PRMU2012-73,HIP2012-62
Volume (vol) vol.112
Number (no) 357
Page pp.pp.-
#Pages 6
Date of Issue