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

Improved Elastic Stroke Matching in Online Handwriting Recognition

IP.com Disclosure Number: IPCOM000119556D
Original Publication Date: 1991-Feb-01
Included in the Prior Art Database: 2005-Apr-01
Document File: 2 page(s) / 89K

Publishing Venue

IBM

Related People

Chefalas, TE: AUTHOR [+2]

Abstract

Disclosed are methods that improve the matching of sloppy writing avoiding the problem of an unknown stroke being out of warp range of a corresponding prototype stroke. A stroke is the writing from pen-down to pen-up.

This text was extracted from an ASCII text file.
This is the abbreviated version, containing approximately 52% of the total text.

Improved Elastic Stroke Matching in Online Handwriting Recognition

      Disclosed are methods that improve the matching of sloppy
writing avoiding the problem of an unknown stroke being out of warp
range of a corresponding prototype stroke.  A stroke is the writing
from pen-down to pen-up.

      A common approach for online recognition of handwritten
characters is to elastically match an unknown character against
prototype characters (1-5).  Many elastic matching algorithms permit
stretching (elasticity) of the prototype from one-half and two times
its original length (3,4,5).  A problem arises when an unknown stroke
is not within the warp range (elasticity range) of its corresponding
prototype.

      The main method disclosed is to use linear matching [6] as a
backup when a stroke is not within warp range for elastic matching.
Since linear matching does not have an elasticity constraint, this
avoids the out-of-warp-range problem.  Also, linear matching precedes
elastic matching (6) so this strategy does not increase computation.
This method decreased the error rate on a set of handwriting data by
15 percent.

      To further improve accuracy, special procedures are used for
short strokes.  After filtering, stroke k is a sequence of nk
coordinates,                          .  For elastic matching, each
point of a stroke is represented by a set of parameters.  Currently,
the parameters used are the direction angle,   , from one point to
the next, and the coordinate values of the point normalized by the
stroke center of gravity, x and y (4).  Thus, the representation of a
stroke for elastic matching becomes

                            (Image Omitted)

      In the improved representation, a stroke consisting of one
point (after filtering) is represented as three points, with
direction angles of 0, 120, and 240 degrees (from a 360-degree
circle), that is where x and y are the normalized coordinates of the
dot. This representation is an improvement for several reasons.
First, true dots are significant events in the recognition process,
and this representation gives a dot significant weight because
strokes are weighted by their number of points in the...