スケーラブルなオンライン・オフライン統合文字認識システムの提案と評価 [in Japanese] A scalable on-line and off-line combined recognition system [in Japanese]
-
- 織田 英人 ODA Hideto
- 東京農工大学 大学院 工学教育部 Graduate School of Technology, Tokyo University of Agriculture and Technology
-
- 耒代 誠仁 KITADAI Akihito
- 東京農工大学 大学院 大学院共生科学技術研究部 Institute of Symbiotic Science and Technology, Tokyo University of Agriculture and Technology
-
- 小沼 元輝 [他] ONUMA Motoki
- 東京農工大学 大学院 工学教育部 Graduate School of Technology, Tokyo University of Agriculture and Technology
-
- 朱 碧蘭 ZHU Bilan
- 東京農工大学 大学院 工学教育部 Graduate School of Technology, Tokyo University of Agriculture and Technology
-
- 中川 正樹 NAKAGAWA Masaki
- 東京農工大学 大学院 大学院共生科学技術研究部 Institute of Symbiotic Science and Technology, Tokyo University of Agriculture and Technology
Search this Article
Author(s)
-
- 織田 英人 ODA Hideto
- 東京農工大学 大学院 工学教育部 Graduate School of Technology, Tokyo University of Agriculture and Technology
-
- 耒代 誠仁 KITADAI Akihito
- 東京農工大学 大学院 大学院共生科学技術研究部 Institute of Symbiotic Science and Technology, Tokyo University of Agriculture and Technology
-
- 小沼 元輝 [他] ONUMA Motoki
- 東京農工大学 大学院 工学教育部 Graduate School of Technology, Tokyo University of Agriculture and Technology
-
- 朱 碧蘭 ZHU Bilan
- 東京農工大学 大学院 工学教育部 Graduate School of Technology, Tokyo University of Agriculture and Technology
-
- 中川 正樹 NAKAGAWA Masaki
- 東京農工大学 大学院 大学院共生科学技術研究部 Institute of Symbiotic Science and Technology, Tokyo University of Agriculture and Technology
Abstract
本稿では,認識システムが出力する評価値を正規化する手法を用いて,オンライン手書き文字認識手法とオフライン手書き文字認識手法を統合する認識システムについて述べる.実現においては,オンライン手法の利点とオフライン手法の利点の融合を指針とした.統合の結果,オンライン手書き文字パターンデータベース「kuchibue」による実験では,150Kバイトのプロトタイプ辞書によるオンライン認識システム単独で87.8%,91Mバイトのプロトタイプ辞書によるオフライン認識システム単独で86.1%の1位認識率に対し,統合認識システムでは,91.3%の1位認識率,98.1%の5位累積認識率を達成した.また,オフラインプロトタイプ辞書を約1Mバイトに圧縮した場合,オフライン認識システム単独で74.2%の1位認識率に対し,同一サイズのオンライン認識システムとの統合認識システムでは1位認識率88.1%,5位累積認識率97.7%を達成した.
This paper describes an on-line/off-line combined recognition system using recognition score normalization. The goal of the combination is to exploit the merits of on-line recognition and off-line recognition. The combined system achieves 91.3% recognition rate and 98.1% 5th-cumulative rate to the on-line handwritten character pattern database "kuchibue" while the on-line recognizer using the 150KB prototype dictionary and off-line recognizer using the 91MB prototype dictionary perform 87.8% and 86.1% recognition, respectively. When the off-line recognizer with the reduced prototype dictionary of 1Mbyte is combined with the above on-line recognizer, the combined system achieves 88.1% recognition rate and 97.7% 5th-cumulative rate, though the reduced off-line recognizer alone performs 74.2% recognition.
Journal
-
- IEICE technical report
-
IEICE technical report 105(615), 135-140, 2006-02-24
The Institute of Electronics, Information and Communication Engineers
References: 19
-
1
- On-line recognition of freely handwritten Japanese characters using directional feature densities
-
KAWAMURA A.
Proc. 11th ICPR, Aug, 1992 II, 183-186, 1992
Cited by (1)
-
2
- On-line Japanese character recognition experiments by an off-line method
-
HAMANAKA M.
Proc. 2nd ICDAR, Oct. 1993, 1993
Cited by (1)
-
3
- Decision Combination in Multiple Classifiers Systems
-
HO T. K.
IEEE Trans. PAMI 16(1), 66-74, 1994
Cited by (1)
-
4
- A Probabilistic Framework for Combining Multiple Classifiers at Abstract Level
-
KANG Hee-Joong
Proc. 4th ICDAR, Aug. 1997, 1997
Cited by (1)
-
5
- Hybrid Pen-Input Character Recognition System Based on Integration of Online-Offline Recognition
-
TANAKA H.
5th ICDAR, Sept, 1999, 1999
Cited by (1)
-
6
- A New Warping Technique for Normalizing Likelihood of Multiple Classifiers and its Effectiveness in Combined On-Line/Off-Line Japanese Character Recognition
-
VELEK O.
Proc. 8th IWFHR, Aug, 2002, 2002
Cited by (1)
-
7
- On-line Handwriting Character Recongnition using Directional Features and Direction-Change Features [in Japanese]
-
OKAMOTO Masayoshi , YAMAMOTO Kazuhiko
The Transactions of the Institute of Electrical Engineers of Japan. C 119(3), 358-366, 1999-03
References (23) Cited by (7)
-
8
- Combining the Classification Results of Independent Classifiers Based on the Dempster/Shafer Theory of Evidence
-
MANDLER E.
Pattern Recognition and Artificial Intelligence, 1988
Cited by (5)
-
9
- A method of combining multiple experts for recognition of unconstrained handwritten numerals
-
HUANG Y. S.
IEEE Trans. PAMI 17(1), 90-94, 1995
Cited by (8)
-
10
- On combining classifiers
-
KITTLER J.
Trans. on PAMI 20(3), 226-239, 1998
DOI Cited by (34)
-
11
- Bagging predictors
-
BREIMAN L.
Machine Learning 24, 123-140, 1996
DOI Cited by (79)
-
12
- A decision-theoretic generalization of on-line learning and an application to boosting
-
FREUND Y.
Journal of Computer and System Sciences 55, 119-139, 1997
DOI Cited by (210)
-
13
- The random subspace method for constructing decision forests
-
HO T. K.
IEEE Transactions on Pattern Analysis and Machine Intelligence 20(8), 832-844, 1998
Cited by (15)
-
14
- On-line recognition of chinese characters : The state of the art
-
LIU C. L.
IEEE Trans. Pattern Anal. Mach. Intell. 26(2), 198-213, 2004
Cited by (15)
-
15
- Modified quadratic discriminant functions and the application to Chinese character recognition
-
KIMURA F.
IEEE Trans. Pattern Anal. Mach. Intell., 1987 PAMI-9(1), 149-153, 1987
Cited by (56)
-
16
- Collection of on-line handwritten Japanese character pattern databases and their analysis
-
NAKAGAWA M.
Int. J. Doc. Anal. Recognit. 7(1), 69-81, 2004
DOI Cited by (8)
-
17
- A Model of On-line Handwritten Japanese Text Recognition Free from Line Direction and Writing Format Constraints
-
NAKAGAWA Masaki , ZHU Bilan , ONUMA Motoki
IEICE Trans. Inf&Syst., D 88(8), 1815-1822, 2005-08-01
References (11) Cited by (11)
-
18
- A Linear-Time Elastic Matching Algorithm for On-line Recognition of Handwritten Japanese Characters [in Japanese]
-
AKIYAMA Katsuhiko , NAKAGAWA Masaki
The Transactions of the Institute of Electronics,Information and Communication Engineers. 00081(00004), 651-659, 1998-04-25
References (20) Cited by (48)
-
19
- Natural Language Processing in Text Recognition [in Japanese]
-
NISHINO Fumihito
IPSJ Magazine 34(10), 1274-1280, 1993-10-15
IPSJ References (53) Cited by (31)
Cited by: 1
-
1
- Development of cursive character pattern recognition for accessing a digital dictionary to support decoding of historical documents [in Japanese]
-
ONUMA Motoki , ZHU Bilan , YAMADA Shoji , SHIBAYAMA Mamoru , NAKAGAWA Masaki
IEICE technical report 106(606), 91-96, 2007-03-09
References (10) Cited by (2)