Invariant handwritten Chinese character recognition

被引：0

作者：

Liu, JNK ^{[1
]}

Lee, RST ^{[1
]}

机构：

[1] Hong Kong Polytech Univ, Dept Comp, Hong Kong, Peoples R China

来源：

ICONIP'98: THE FIFTH INTERNATIONAL CONFERENCE ON NEURAL INFORMATION PROCESSING JOINTLY WITH JNNS'98: THE 1998 ANNUAL CONFERENCE OF THE JAPANESE NEURAL NETWORK SOCIETY - PROCEEDINGS, VOLS 1-3 | 1998年

关键词：

Dynamic Link Architecture; genetic algorithm; neural networks;

D O I：

暂无

中图分类号：

TP18 [人工智能理论];

学科分类号：

081104 ; 0812 ; 0835 ; 1405 ;

摘要：

A model based on the hybridization of genetic algorithm (GA) with dynamic link Architecture is presented for the off-line handwritten Chinese character recognition. Using a pattern sample of 100 handwritten Chinese characters with 400 handwritten testing samples for various invariant tests including translation, rotation, reflection, and distortion, an overall 20% improvement of correct recognition rate has been attained, giving a correct recognition rate has been attained, giving a correct recognition rate of 90% for the translation transformation.

引用

页码：275 / 278

页数：4

共 4 条

[1] A NEURAL NETWORK FOR INVARIANT PATTERN-RECOGNITION
BIENENSTOCK, E
VONDERMALSBURG, C
[J]. EUROPHYSICS LETTERS, 1987, 4 (01): : 121 - 126
[2] A discrete version of the dynamic link network
Feng, JF
Tirozzi, B
[J]. NEUROCOMPUTING, 1997, 15 (02) : 91 - 106
[3] Invariant character recognition in dynamic link architecture
Liu, JNK
Lee, RST
[J]. 1997 IEEE KNOWLEDGE AND DATA ENGINEERING EXCHANGE WORKSHOP, PROCEEDINGS, 1997, : 188 - 195
[4] PATTERN-RECOGNITION BY LABELED GRAPH MATCHING
VONDERMALSBURG, C
[J]. NEURAL NETWORKS, 1988, 1 (02) : 141 - 148

← 1 →