Browse Prior Art Database

Matching Algorithm for On-Line Character Recognition

IP.com Disclosure Number: IPCOM000104403D
Original Publication Date: 1993-Apr-01
Included in the Prior Art Database: 2005-Mar-19
Document File: 4 page(s) / 94K

Publishing Venue

IBM

Related People

Kaneko, H: AUTHOR [+4]

Abstract

Disclosed is a matching algorithm for on-line character recognition that requires only a small dictionary because, instead of variable weighting factors, it uses ordered constant ones proportional to the frequency of appearance of the stroke type.

This text was extracted from an ASCII text file.
This is the abbreviated version, containing approximately 53% of the total text.

Matching Algorithm for On-Line Character Recognition

      Disclosed is a matching algorithm for on-line character
recognition that requires only a small dictionary because, instead of
variable weighting factors, it uses ordered constant ones
proportional to the frequency of appearance of the stroke type.

      The following description is based on the example of the
Chinese character (See Table 1, which in Japanese is read 'Migi' and
means 'right-hand').  We observed twenty-four examples of this
character and extracted statistical features.  The six defined types
(A to F) of elementary strokes are shown in Fig. 1, and the
combinations of stroke patterns of samples are listed in Table 1.
There are two main types of 'Migi' samples: those written in five
strokes and those written in four strokes.  Internal expressions for
this character in the dictionary are shown in Table 2.

      Every input stroke is first categorized as the nearest of the
elementary strokes A to F by using the DP matching method.  An
elementary stroke string such as 'CABDA' is then generated from the
stroke codes, and this string is compared with the dictionary.  An
example of a comparison method is to add similarity points when the
elementary stroke codes of the ith stroke of the input and the
dictionary are formed to match.  Candidates are then sorted in order
of similarity points, and the top candidate is output as the
recognition result.

      Table 3 shows statistical similar...