An algorithm to recognize isolated, overlapped and connected handwritten numerical characters

Author

Flores, Edna L. ; Rezende, Mer N. ; Carrijo, Gilberto A. ; Yabu-uti, Joao B T

Author_Institution

Dept. de Engenharia Eletrica, Univ. Fed. de Uberlandia, Brazil

fYear

1995

fDate

35030

Firstpage

53

Lastpage

58

Abstract

The authors describe the results of their investigation of the development of an algorithm for recognizing and identifying handwritten numerical characters which may be isolated, overlapped and connected. These numerals may be continuous or fragmented. The features that the digits have in common and the intrinsic features of each one were considered. Such characteristics are used in the human recognition process and are the basis of the proposed algorithm. The principal idea of the algorithm is to start at the top of the image and group the pixels-on in horizontal and vertical lines that will characterize each one of the digits. Numerous experimental tests were carried out for recognition of numerical characters of bank checks and an accuracy of 99.88% was obtained

Keywords

handwriting recognition; image recognition; word processing; accuracy; bank checks; connected handwritten numerical character recognition; digit recognition; experimental tests; human recognition process; intrinsic features; pixels-on; Character recognition; Handwriting recognition; Humans; Pixel; Testing;

fLanguage

English

Publisher

ieee

Conference_Titel

Intelligent Information Systems, 1995. ANZIIS-95. Proceedings of the Third Australian and New Zealand Conference on

Conference_Location

Perth, WA

Print_ISBN

0-86422-430-3

Type

conf

DOI

10.1109/ANZIIS.1995.705714

Filename

705714