Image analysis – Pattern recognition – On-line recognition of handwritten characters
Reexamination Certificate
1998-06-04
2001-05-01
Couso, Jose L. (Department: 2721)
Image analysis
Pattern recognition
On-line recognition of handwritten characters
C382S181000, C382S188000, C382S289000, C382S290000
Reexamination Certificate
active
06226404
ABSTRACT:
BACKGROUND OF THE INVENTION
The present invention relates to an on-line character recognition system for recognizing characters input by handwriting, making use of a pen-based input device.
In many conventional on-line character recognition systems, the character recognition is performed by examining matches of input character patterns to reference character patterns prepared beforehand. Therefore, it is required for precise recognition that the direction of the input character patterns not be different from the direction of the reference character patterns. However, when the input character patterns are hand-written, especially making use of a small tablet of an electronic pocket book, for example, the input character patterns may often be written by being inclined against the input frame or the tablet, a factor which degrades recognition performance.
To overcome this problem, some devices for normalizing the direction of the inclined character patterns have been proposed, as disclosed in Japanese patent application laid open as a Provisional Publication No. 269265/'88 (hereafter called the first prior art) or Japanese patent application laid open as a Provisional Publication No. 337963/'94 (hereafter called the second prior art).
In the first prior art, a straight line
10
, as illustrated in
FIG. 6
, is written by a user before the user begins to input characters for indicating his preferred pattern inclination to be hand-written. Along the straight line
10
, the user writes an intended pattern
9
of a character ‘A’, for example. By normalizing the pattern direction, referring the inclination of the straight line
10
, an on-line character recognition system of the first prior art performs the pattern matching of the intended pattern
9
.
For indicating the preferred pattern inclination, a rotary switch
11
of
FIG. 7
is also proposed in the first prior art. The user manipulates the rotary switch
11
according to the inclination of the handwritten pattern
9
instead of writing the straight line
10
. The on-line character recognition system of the first prior art normalizes the pattern direction referring to the rotation angle of the rotary switch
11
.
In the second prior art, which is intended to be applied when two or more users facing each other input handwriting characters using a single tablet, for example, the direction of an input pattern are discriminated according to where the input pattern is written in the tablet or by which stylus pen the input pattern is written, by preparing two or more stylus pens. For each input field, or each stylus pen, a character direction is preset and the coordinates of the input pattern are normalized according to the character direction. The character direction may be discriminated making use of information of a beginning point of character strings or an input mode designated by the user.
However, with the inclination normalization method of the first prior art, the user is forced to write the straight line
10
or to manipulate the rotary switch
11
for indicating the pattern direction, which is burdensome for the user.
According to the second prior art, the pattern inclination cannot be normalized but with the character direction preset for each input field of the stylus pen. Therefore, when the character pattern is written differently to the preset character direction, it may not be recognized correctly, which is the same when the character direction is discriminated making use of other information.
SUMMARY OF THE INVENTION
Therefore, a primary object of the present invention is to provide an on-line character recognition system which can recognize handwritten character patterns correctly by normalizing inclination of input character patterns precisely, with little additional burden charged to the user, by making use of information of inclination angle of a stylus pen itself used for writing the character patterns.
In order to achieve the object, an on-line character recognition system of the invention comprises;
a pen-based input device for outputting data of a character pattern hand-written with a stylus pen on an input plane of the pen-based input device, and outputting inclination angle information of the stylus pen in reference to the input plane,
a slant angle detector for obtaining a slant angle of the stylus pen from the inclination angle information, said slant angle representing an angle of vertical projection of the stylus pen on the input plane;
a slant angle register for storing standard slant angle data of the stylus pen,
a slant compensation section for outputting compensated data set by compensating the data of the character pattern output from the pen-based input device making use of the slant angle obtained by the slant angle detector and the standard slant angle data registered in the slant angle register, and
a character recognition section for recognizing a character represented by the compensated data set output from the slant compensation section.
There is a substantial correspondence between the direction of a character pattern and the slant angle of the stylus pen with which the character pattern is written. When the slant angle changes, the direction of the character pattern changes substantially with the same degree. Therefore, the inclination of input character patterns can be normalized correctly and a precise character recognition result can be obtained according to the invention without any additional burden charged to the user.
As to the standard slant angle data, it may be pre-determined according to a statistical examination, or it may be obtained from output of the slant angle detector corresponding to the data of a character hand-written by a user for a learning procedure.
The standard slant angle data may be classified for each writing style classification of users, providing a selector for selecting appropriate standard slant angle data to be referred to by the slant compensation section from the standard slant angle data according to a writing style of one of the users.
REFERENCES:
patent: 4495646 (1985-01-01), Gharachorloo
patent: 4947156 (1990-08-01), Sato et al.
patent: 5101439 (1992-03-01), Kiang
patent: 5333209 (1994-07-01), Sinden et al.
patent: 5453761 (1995-09-01), Tanaka
patent: 5544260 (1996-08-01), Chefalas et al.
patent: 5588073 (1996-12-01), Lee et al.
patent: 5615284 (1997-03-01), Rhyne et al.
patent: 5633954 (1997-05-01), Gupta et al.
patent: 5696538 (1997-12-01), Kishi
patent: 5740273 (1998-04-01), Parthasarathy et al.
patent: 5835632 (1998-11-01), Takasu et al.
patent: 5903668 (1999-05-01), Beernink
patent: 5966460 (1999-10-01), Porter, III et al.
patent: 5981884 (1999-11-01), Sato et al.
patent: 2-201520 (1990-08-01), None
patent: 3-288173 (1991-12-01), None
patent: 6-19614 (1994-01-01), None
patent: 6-150067 (1994-05-01), None
patent: 6-258914 (1994-09-01), None
patent: 6-337963 (1994-12-01), None
patent: 63-269265 (1998-11-01), None
Beigi, et al “Size Normalzation in On-Line Unconstrained Handwriting Recognition”, IEEE, pp. 173, 1994.*
Kobayashi, et al “An On-Line Handwriting Character Recognition Algorithm RAV (Reparameterized Angle Variations)”, IEEE, pp. 919-925, Apr. 1997.*
“WACOM UD-Series Digitizer”, produced by WACOM, a corporation of Japan, Dec. 20, 1995.
M. Hamanaka, et al., “On-Line Japanese Character Recognition Experiments by an Off-Line Method”, Proceeding of the Second International Conference on Document Analysis and Recognition, Oct. 1993, pp. 204-207.
Couso Jose L.
Mariam Daniel G.
McGinn & Gibb PLLC
NEC Corporation
LandOfFree
On-line character recognition system does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with On-line character recognition system, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and On-line character recognition system will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-2538461