Direction-Change Features of Imaginary Strokes
for On-line Handwriting Character Recognition
Masayoshi Okamoto
*
Akira Nakamura
*
*
Hypermedia Research Center,
SANYO Electric Co.,Ltd
180,Ohmori,Anpachi-Cho, Gifu, 503-0195 Japan
okamo@gf.hm.rd.sanyo.co.jp,
naka@gf.hm.rd.sanyo.co.jp
Kazuhiko Yamamoto
+
+
Faculty of Engineering,
Gifu University
1-1 Yanagido, Gifu 501-1193 Japan
yamamoto@info.gifu-u.ac.jp
Abstract
We found suitable direction-change features of the imaginary
strokes in the pen-up state for on-line handwritten cursive
character recognition. Our method simultaneously uses both
directional features, otherwise known as off-line features, and
direction-change features, which we designed as on-line features.
The directional features express where and in which direction
each character’s coordinates exist. The direction-change features
express where and in which direction each direction of the
character’s coordinates change, and express where the circular
parts of the character exist. These direction-change features
express both written strokes in the pen-down state and unwritten
imaginary strokes in the pen-up state. It is important to get
suitable direction-change features when using this method. We
tried to examine the influence on character recognition rates
when changing the functions used to get each direction-change
feature based on the imaginary stroke lengths. Then, we found
that the best function is the function which puts no weight on the
imaginary stroke lengths. The recognition rate for freely-written
Japanese characters was improved from 82.37% to 86.32 % by
our new method using the best function as opposed to our old
method using a function which gets each direction change feature
in inverse proportion to the imaginary stroke lengths.
1.
Introduction
On-line character recognition technology is important for
human-interfaces. Many character recognition methods have
been researched[1]. Most of these methods are based on stroke
matching[2]. Currently, if you write Japanese characters neatly,
these characters are recognized correctly. However, it is difficult
to recognize cursive handwritten characters with stroke-number
and stroke-order variations. Recently, some recognition methods
for cursive handwritten characters have been researched. Typical
methods are the Wakahara Method using a stroke-based Affine
transformation[3], the Nakagawa Method with customizable
recognition[4], which improves the recognition rate using pure
online features, and the Hamanaka-Yamada-Tsukumo
Method[5][6] with directional pattern matching[7-9] which uses
off-line features. We proposed a new handwritten character
recognition method[13], called DDCPM (Directional and
Direction-Change Pattern Matching), simultaneously using
directional features[7-9] and direction-change features which we
designed as on-line features. We showed that our methods using
both directional features and direction-change features was able to
obtain higher recognition rates than the traditional method using
only directional features. We showed that the recognition rate was
further improved when direction-change features take into
account unwritten imaginary strokes in the pen-up state.
In this paper, we explain suitable direction-change features of
the imaginary strokes in the pen-up state. We tried to examine the
influences to character recognition rates when changing the
functions used to get each direction-change feature based on the
imaginary stroke lengths, then, we found the best function.
2. Recognition Method
In this section, we explain our recognition process [13]. First,
on-line character data ( x,y coordinate data) is transformed to
bitmap data. Next, the bitmap data and on-line data are
nonlinearly normalized by Line Density Equalization[10][11].
Directional features are then extracted from the normalized
bitmap data, and direction-change features are extracted from the
normalized on-line data. After these feature patterns are blurred,
the patterns’ dimensions are reduced. The reduced dimensional
feature patterns of the inputted characters are compared with the
reduced dimensional feature patterns of the standard characters,