Dismiss
InnovationQ will be updated on Sunday, Oct. 22, from 10am ET - noon. You may experience brief service interruptions during that time.
Browse Prior Art Database

Global Prototype Establishment Procedure for Handwritten Character Recognition

IP.com Disclosure Number: IPCOM000034340D
Original Publication Date: 1989-Feb-01
Included in the Prior Art Database: 2005-Jan-27
Document File: 4 page(s) / 40K

Publishing Venue

IBM

Related People

Fox, AS: AUTHOR [+2]

Abstract

A technique is described whereby a global prototype establishment procedure for handwritten character recognition examines the inter- relationships among a set of candidates, choosing an appropriate set. The concept is an improvement over previous methods in that a global approach is used to choose an appropriate subset to become prototypes, rather than using a serial approach in character examination. Typically, there are three essential criteria used in the choosing of a subset: 1) There should be at least one prototype for each distinct way of writing a character. 2) There should be provision for "prioritizing" the selection of prototypes. Generally, priority should be given to those characters written most naturally, in terms of shape and speed of writing.

This text was extracted from a PDF file.
At least one non-text object (such as an image or picture) has been suppressed.
This is the abbreviated version, containing approximately 47% of the total text.

Page 1 of 4

Global Prototype Establishment Procedure for Handwritten Character Recognition

A technique is described whereby a global prototype establishment procedure for handwritten character recognition examines the inter- relationships among a set of candidates, choosing an appropriate set. The concept is an improvement over previous methods in that a global approach is used to choose an appropriate subset to become prototypes, rather than using a serial approach in character examination. Typically, there are three essential criteria used in the choosing of a subset: 1) There should be at least one prototype for each distinct

way of writing a character.

2) There should be provision for "prioritizing" the

selection of prototypes. Generally, priority should be

given to those characters written most naturally, in

terms of shape and speed of writing. For example, in

the recognition of run-on discrete characters, priority

should be given to characters written in the run-on mode,

rather than to characters written discretely. Priority

may also be given to characters written recently in time

and/or to those written rapidly.

3) Where possible, "mavericks" should not be chosen as

prototypes. A "maverick" is a character that is not

written as intended, inadvertently not made differently

from the other alphabet characters, not captured

accurately due to hardware problems, or distorted in some

fashion due to other problems. This criteria is

difficult to meet without feedback from the writer to

identify the "mavericks"; for example, even pen skips

cannot automatically be identified unequivocally. Even though "mavericks" cannot be identified unequivocally without feedback to the writer, given a sufficiently large set of candidate prototypes, they can be detected automatically with reasonable accuracy by examining the distance relationships among the candidate prototypes. For example, if a writer usually made the number '0' differently from the letter 'O', perhaps by slashing the character for the letter O, an unslashed letter O would be closer to the number 0 and could be identified as a potential "maverick". This information could then be presented to the writer, or, in some cases, can be acted on automatically without writer interaction. Automatic operation is particularly important for applications, such as filling out coding sheets, where raw accuracy must be maximized because redundant contextual information is not available. "Mavericks", which are inadvertently not made differently from the other alphabet characters can be handled by the use of contextual information. For example, the letter 'I' and the number '1' could both have similar single stroke prototypes, but may be distinguished by means of context. However, until contextual algorithms become available, such characters will be considered "mavericks". Selecting the best appropriate subset may require some interaction and/or verification

1

Page 2 of 4

from the writer and perhaps also from a...