|
Accession number;99A0938072
|
| Title;A Study on Accurate Character Recognition Considering the Distribution of Feature Vectors. |
| Author;
SUN F
(Tohoku Univ.)
|
Journal Title;Record of Electrical and Communication Engineering Conversazione, Tohoku University
|
Journal Code:F0511A
|
ISSN:0385-7719
|
|
VOL.68;NO.1;PAGE.28-31(1999)
|
| Figure&Table&Reference;REF.6 |
| Pub. Country;Japan |
| Language;Japanese |
| Abstract;Dictionary construction and discriminant function are very important factors of character recognition system. In this paper, by considering the distribution of feature vectors, new algorithms to construct a multi-template dictionary for character recognition are presented. Moreover, Simplified Mahalanobis distance and Divided Mahalanobis distance are proposed as the discriminant functions that can approximate the Mahalanobis distance in the case of small size of training samples. The effectiveness of these two functions are investigated by simulations. The effectiveness of the proposed methods based on these disriminant functions are shown with the experimental results on ETL9B. (author abst.) |
|
|
|
Related Articles;
|
