Browse Prior Art Database

Learning Method for On-Line Character Recognizers Based on User's Heuristic Data

IP.com Disclosure Number: IPCOM000116009D
Original Publication Date: 1995-Jul-01
Included in the Prior Art Database: 2005-Mar-30
Document File: 2 page(s) / 51K

Publishing Venue

IBM

Related People

Souya, T: AUTHOR

Abstract

Disclosed is for offering a learning method based on users' heuristic data for on-line character recognizers. The characteristic features of the method are as follows: 1. Users can provide information to a recognizer about which parts of a newly registered character pattern have significant features (these parts are called "heuristic parts" (Fig. 1). 2. The information on heuristic parts is used selectively in recognition processes (Fig. 2) to eliminate confusion associated with newly registered patterns.

This text was extracted from an ASCII text file.
This is the abbreviated version, containing approximately 86% of the total text.

Learning Method for On-Line Character Recognizers Based on User's
Heuristic Data

      Disclosed is for offering a learning method based on users'
heuristic data for on-line character recognizers.  The characteristic
features of the method are as follows:
  1.  Users can provide information to a recognizer about which parts
       of a newly registered character pattern have significant
features
       (these parts are called "heuristic parts" (Fig. 1).
  2.  The information on heuristic parts is used selectively in
       recognition processes (Fig. 2) to eliminate confusion
associated
       with newly registered patterns.

      A new pattern to be registered is selected without the support
of any of the statistics that are normally used to create character
pattern dictionaries.  Thus, confusion often occurs between a new
pattern and known patterns already included in the character pattern
dictionaries.

      If a confusion is predetected, two common ways to prevent it
are to cancel the registration itself and to purge the known patterns
that cause confusion.  However, these ways reduce the efficiency of
pattern registration.

      Consequently, it is necessary to find a way of eliminating
confusion without the use of any statistics.

      The main purpose of the invention is to eliminate the confusion
associated with pattern registration.  The disclosed method will
include features for eliminating such confusion bas...