Browse Prior Art Database

Gesture Recognition with Shape Matcher for Handwriting Recognition

IP.com Disclosure Number: IPCOM000105011D
Original Publication Date: 1993-Jun-01
Included in the Prior Art Database: 2005-Mar-19
Document File: 2 page(s) / 82K

Publishing Venue

IBM

Related People

Chefalas, T: AUTHOR [+3]

Abstract

GESTURE and HANDWRITING - Pen based computers use hand drawn gestures and handwriting for operation and data entry[1]. Handwriting is usually regular in size and shape while gesture varies more in size and shape. To recognize both with high accuracy, it was customary to have two recognizers[2,3], each tuned for respective vocabulary. This approach works well, but requires a lot of memory, which is at a premium in a portable, battery-operated computing environment.

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

Gesture Recognition with Shape Matcher for Handwriting Recognition

      GESTURE and HANDWRITING - Pen based computers use hand drawn
gestures and handwriting for operation and data entry[1].
Handwriting is usually regular in size and shape while gesture varies
more in size and shape.  To recognize both with high accuracy, it was
customary to have two recognizers[2,3], each tuned for respective
vocabulary.  This approach works well, but requires a lot of memory,
which is at a premium in a portable, battery-operated computing
environment.

      In this disclusure, an existing, shape recognizer for
handwriting is utilized to recognize gestures with variations, thus
saving valuable memory.  By judicious pre- and post- processing, it
is possible to achieve good accuracy and speed.

GESTURAL VARIATIONS - There are more variations in different
instances of gestures than in handwriting.  Major variations are in
size, and aspect ratio.  On the other hand, gestures tend to be more
sensitive to stroke order and direction than handwriting.  Shape
matching developed for handwriting forgives some variations, but
gestures exhibit variations more than the matcher can handle.
Pre-processing stage is designed to normalize the gestures such that
variations are reduced down to matchers range.

PRE-PROCESSING - Unlike handwriting, which sits on a baseline and
fixed character height, gestures can be drawn anywhere on the screen
in different sizes.  The first step is to pre-process the input
strokes for later match process.

      A rectangular box (square works well) is drawn around the input
strokes where the bottom of the box becomes the baseline and height
of the box becomes the character height for matching purposes.  Then
the box is scaled to the size of the pre-stored prototypes and shape
matching is performed.

      Care must be taken in drawing the rectangular box that a dot
does not turn into a line, and slightly mis-oriented horizontal line
does not turn into a slash.  The box must have a minimum dimension,
and fixed aspect ratio (Figure).

HOT POINT - The shape matcher returns a prototype id of the best
match.  For handwriting, this is sufficient, but for gestures,
additional information is necessary.  A gesture must signify an
operat...