Radical region based CNN for offline handwritten Chinese character recognition

Luo Weike, Seiichiro Kamata

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Abstract

In recent years, deep learning based methods have been widely used in handwritten Chinese character recognition (HCCR) and greatly improved the recognition accuracy. However, most of the current methods simply employ famous networks like GoogleNet without fully embedding the specific features of Chinese characters. Taking structural characteristics into consideration, we propose a radical region network structure to represent the radical region information (For example left, right, top and bottom radical regions). In our study, the character feature is represented as global feature while the radical region feature is represented as local feature. The multi-supervised training method is also used to learn two kinds of feature at the same time. Experiment results show the proposed methods improve recognition accuracy of current models. The performance of the best model has been raised to 97.42% on ICDAR 2013 offline HCCR competition database which achieves the state-of-The-Art result as we know.

Original languageEnglish
Title of host publicationProceedings - 4th Asian Conference on Pattern Recognition, ACPR 2017
PublisherInstitute of Electrical and Electronics Engineers Inc.
Pages548-553
Number of pages6
ISBN (Electronic)9781538633540
DOIs
Publication statusPublished - 2018 Dec 13
Event4th Asian Conference on Pattern Recognition, ACPR 2017 - Nanjing, China
Duration: 2017 Nov 262017 Nov 29

Publication series

NameProceedings - 4th Asian Conference on Pattern Recognition, ACPR 2017

Other

Other4th Asian Conference on Pattern Recognition, ACPR 2017
CountryChina
CityNanjing
Period17/11/2617/11/29

Fingerprint

Character recognition
Experiments
Deep learning

Keywords

  • Deep learning
  • Offline handwritten Chinese character recognition
  • Radical region information

ASJC Scopus subject areas

  • Artificial Intelligence
  • Computer Vision and Pattern Recognition
  • Signal Processing

Cite this

Weike, L., & Kamata, S. (2018). Radical region based CNN for offline handwritten Chinese character recognition. In Proceedings - 4th Asian Conference on Pattern Recognition, ACPR 2017 (pp. 548-553). [8575881] (Proceedings - 4th Asian Conference on Pattern Recognition, ACPR 2017). Institute of Electrical and Electronics Engineers Inc.. https://doi.org/10.1109/ACPR.2017.76

Radical region based CNN for offline handwritten Chinese character recognition. / Weike, Luo; Kamata, Seiichiro.

Proceedings - 4th Asian Conference on Pattern Recognition, ACPR 2017. Institute of Electrical and Electronics Engineers Inc., 2018. p. 548-553 8575881 (Proceedings - 4th Asian Conference on Pattern Recognition, ACPR 2017).

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Weike, L & Kamata, S 2018, Radical region based CNN for offline handwritten Chinese character recognition. in Proceedings - 4th Asian Conference on Pattern Recognition, ACPR 2017., 8575881, Proceedings - 4th Asian Conference on Pattern Recognition, ACPR 2017, Institute of Electrical and Electronics Engineers Inc., pp. 548-553, 4th Asian Conference on Pattern Recognition, ACPR 2017, Nanjing, China, 17/11/26. https://doi.org/10.1109/ACPR.2017.76
Weike L, Kamata S. Radical region based CNN for offline handwritten Chinese character recognition. In Proceedings - 4th Asian Conference on Pattern Recognition, ACPR 2017. Institute of Electrical and Electronics Engineers Inc. 2018. p. 548-553. 8575881. (Proceedings - 4th Asian Conference on Pattern Recognition, ACPR 2017). https://doi.org/10.1109/ACPR.2017.76
Weike, Luo ; Kamata, Seiichiro. / Radical region based CNN for offline handwritten Chinese character recognition. Proceedings - 4th Asian Conference on Pattern Recognition, ACPR 2017. Institute of Electrical and Electronics Engineers Inc., 2018. pp. 548-553 (Proceedings - 4th Asian Conference on Pattern Recognition, ACPR 2017).
@inproceedings{1eb8abc4042c409aac3513ba426f8f99,
title = "Radical region based CNN for offline handwritten Chinese character recognition",
abstract = "In recent years, deep learning based methods have been widely used in handwritten Chinese character recognition (HCCR) and greatly improved the recognition accuracy. However, most of the current methods simply employ famous networks like GoogleNet without fully embedding the specific features of Chinese characters. Taking structural characteristics into consideration, we propose a radical region network structure to represent the radical region information (For example left, right, top and bottom radical regions). In our study, the character feature is represented as global feature while the radical region feature is represented as local feature. The multi-supervised training method is also used to learn two kinds of feature at the same time. Experiment results show the proposed methods improve recognition accuracy of current models. The performance of the best model has been raised to 97.42{\%} on ICDAR 2013 offline HCCR competition database which achieves the state-of-The-Art result as we know.",
keywords = "Deep learning, Offline handwritten Chinese character recognition, Radical region information",
author = "Luo Weike and Seiichiro Kamata",
year = "2018",
month = "12",
day = "13",
doi = "10.1109/ACPR.2017.76",
language = "English",
series = "Proceedings - 4th Asian Conference on Pattern Recognition, ACPR 2017",
publisher = "Institute of Electrical and Electronics Engineers Inc.",
pages = "548--553",
booktitle = "Proceedings - 4th Asian Conference on Pattern Recognition, ACPR 2017",

}

TY - GEN

T1 - Radical region based CNN for offline handwritten Chinese character recognition

AU - Weike, Luo

AU - Kamata, Seiichiro

PY - 2018/12/13

Y1 - 2018/12/13

N2 - In recent years, deep learning based methods have been widely used in handwritten Chinese character recognition (HCCR) and greatly improved the recognition accuracy. However, most of the current methods simply employ famous networks like GoogleNet without fully embedding the specific features of Chinese characters. Taking structural characteristics into consideration, we propose a radical region network structure to represent the radical region information (For example left, right, top and bottom radical regions). In our study, the character feature is represented as global feature while the radical region feature is represented as local feature. The multi-supervised training method is also used to learn two kinds of feature at the same time. Experiment results show the proposed methods improve recognition accuracy of current models. The performance of the best model has been raised to 97.42% on ICDAR 2013 offline HCCR competition database which achieves the state-of-The-Art result as we know.

AB - In recent years, deep learning based methods have been widely used in handwritten Chinese character recognition (HCCR) and greatly improved the recognition accuracy. However, most of the current methods simply employ famous networks like GoogleNet without fully embedding the specific features of Chinese characters. Taking structural characteristics into consideration, we propose a radical region network structure to represent the radical region information (For example left, right, top and bottom radical regions). In our study, the character feature is represented as global feature while the radical region feature is represented as local feature. The multi-supervised training method is also used to learn two kinds of feature at the same time. Experiment results show the proposed methods improve recognition accuracy of current models. The performance of the best model has been raised to 97.42% on ICDAR 2013 offline HCCR competition database which achieves the state-of-The-Art result as we know.

KW - Deep learning

KW - Offline handwritten Chinese character recognition

KW - Radical region information

UR - http://www.scopus.com/inward/record.url?scp=85060520040&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85060520040&partnerID=8YFLogxK

U2 - 10.1109/ACPR.2017.76

DO - 10.1109/ACPR.2017.76

M3 - Conference contribution

AN - SCOPUS:85060520040

T3 - Proceedings - 4th Asian Conference on Pattern Recognition, ACPR 2017

SP - 548

EP - 553

BT - Proceedings - 4th Asian Conference on Pattern Recognition, ACPR 2017

PB - Institute of Electrical and Electronics Engineers Inc.

ER -