1992 Fiscal Year Final Research Report Summary

Parallel Distributed Models with Inductive Learning Mechanism for Hand-written Characters Recognition

Research Project

Project/Area Number	03650308
Research Category	Grant-in-Aid for General Scientific Research (C)
Allocation Type	Single-year Grants
Research Field	情報工学
Research Institution	Kyushu Institute of Technology
Principal Investigator	EJIMA Toshiaki Kyushu Institute of Technology, 情報工学部, 教授 (00124553)
Co-Investigator(Kenkyū-buntansha)	OHASHI Takeshi Kyushu Institute of Technology, 情報工学部, 助手 (00233239) YOSHIDA Takaichi Kyushu Institute of Technology, 情報工学部, 助教授 (70200996)
Project Period (FY)	1991 – 1992
Keywords	character recognition / neural networks / parallel distributed processing / learning
Research Abstract	First, a character recognition system is developed based on a new neural networks called FPM. FPM is considered as a generalization of so called BP model, which is multi-layer perception proposed by Rumelhart in 1986 and is proved to be useful especially in pattern recognition. Next, to design a robust recognition system against noise, the following factors are considered to be solved: (1) rotation caused by image scanner (2) binary image from multi-gray scale image (3) removal of noise (4) features being robust against distortion (5) segmentation (6) utilization of knowledge about shape of each character Finally, a much improved character recognition system, which is successful in integrating different kinds of techniques concerning above issues, is developed.