Browse Prior Art Database

Optimal Recognition of Baseline Dependent Characters in Online Handwriting Recognition

IP.com Disclosure Number: IPCOM000121287D
Original Publication Date: 1991-Aug-01
Included in the Prior Art Database: 2005-Apr-03
Document File: 2 page(s) / 62K

Publishing Venue

IBM

Related People

Chefalas, TE: AUTHOR [+3]

Abstract

For some alphabets, there are characters that are similar, or even identical, to other characters except for their position relative to the baseline. In English there are several: comma versus single quote, upper versus lowercase P, etc. If the characters are written identically, their discrimination is optimally based solely on the offset from the baseline, and probability distributions for implementing the discrimination can be obtained from known (training) data. Usually, however, the characters are written somewhat differently. Therefore, what is desirable is a formulation that combines both shape and the baseline information.

This text was extracted from an ASCII text file.
This is the abbreviated version, containing approximately 56% of the total text.

Optimal Recognition of Baseline Dependent Characters in Online Handwriting
Recognition

      For some alphabets, there are characters that are
similar, or even identical, to other characters except for their
position relative to the baseline.  In English there are several:
comma versus single quote, upper versus lowercase P, etc.  If the
characters are written identically, their discrimination is optimally
based solely on the offset from the baseline, and probability
distributions for implementing the discrimination can be obtained
from known (training) data.  Usually, however, the characters are
written somewhat differently.  Therefore, what is desirable is a
formulation that combines both shape and the baseline information.

      Recent work on matching by decomposition [1,2] gives the
background for developing such a formulation.  The x component of the
match distance of unknown u against prototype p for a character is
defined as the sum of the squared differences between the normalized
(by the center of gravity) unknown and prototype coordinate values.
This can be expressed as the sum of the squared differences between
the unnormalized unknown and prototype coordinate values, corrected
by a weighted squared difference of the centers of gravity, that is
where n is the number of points in the character.  For simplicity,
computations are presented only for x coordinates; they are similar
to y coordinates.  This formula holds for a linear match of equal
numb...