Since the numerics to be read by the OCR's were mainly used in accounting forms in the business, the requirement to make substitution errors negligibly small was very severe. The main efforts were devoted in order to make substitution errors minimal.
The ANK OCR development project was assigned to be classified, so we could not write any papers on the achievements. After project finished, the research looked very obsolete for us, because several years had already elapsed after the peak of the development. I have noticed recently no papers were published regarding our project which consumed the labor of some hundred man-months. I am very sorry for my colleagues on the infertile academic achievements of the project.
As for the kanji recognition, it is difficult to recognize a sole character
by itself; many pairs cannot be distinguished even by the human beings. So the
AI approach, or post-processing, plays a very important role in kanji
recognition. My effort was related to the high speed post-processing of
recognized results. The work did not end in my Hitachi days, but young
colleagues have made many splendid works after I resigned.
References
[Research Themes of Prof. Nakano] [Prof. Nakano's Welcome Page]
| mail address: | |
<- Please enter the string into the address field |
First Written Before June 17, 1998
Transplanted to KSU June 19, 2003
Transplanted to So-net April 22, 2007
Last Update April 22, 2007
© Yasuaki Nakano 1998-2007