Image analysis – Pattern recognition – On-line recognition of handwritten characters
Reexamination Certificate
1997-05-30
2001-01-16
Couso, Jose L. (Department: 2724)
Image analysis
Pattern recognition
On-line recognition of handwritten characters
C382S185000
Reexamination Certificate
active
06175651
ABSTRACT:
BACKGROUND OF THE INVENTION
1. Field of the Invention
The present invention relates to an on-line character recognition method and device, and more particularly to an on-line character recognition method and device that recognizes inputted characters on-line by finding inter-stroke distance in the patterns in stroke units of inputted characters and reference patterns in stroke units that are produced in advance by quantizing the strokes of target characters.
2. Description of the Related Art
Japanese Patent Publication No. 38351/93 (Hereinbelow referred to as Reference 1) discloses one example of an on-line character recognition device that recognizes characters on-line by using information regarding strokes, a stroke being a continuous pen track from the contact of a pen with a tablet until separation of the pen from the tablet.
In this on-line character recognition device, patterns in stroke units are first found from inputted characters, and then, using a stroke code book composed of assemblages of reference patterns prepared in advance by quantizing the strokes of target characters, inter-stroke distances are found with respect to each of the reference patterns in the stroke code book. Next, inter-character distances are found by adding each of the obtained inter-stroke distances in accordance with a character dictionary in which characters are represented by code strings for each reference pattern within the stroke code book, and the character giving the smallest distance is taken as the recognition result. This method allows a reduction in both the amount of memory and number of calculations as compared with a case in which inter-stroke distances for all strokes are calculated without quantizing the strokes of target characters.
The method described in the paper “Application of Fuzzy Logic in an On-line Handwritten Character Recognition Method” in the Papers of the Institute of Electronics, Information and Communication Engineers (February 1991, D-II, Vol. J74-D-II, No. 2, pp. 166-174) (Hereinafter referred to as Reference 2) seeks the degree of similarity between the shape of each stroke of an inputted character and each of a number of basic stroke shape patterns prepared in advance. Characters are recognized using a recognition dictionary of character units described in terms of fuzzy propositions using this degree of similarity and each of various features. The character dictionary may take a form in which, for example, every character is described in terms of fuzzy propositions using the various features such as stroke shape, stroke length, stroke direction, stroke position, and number of stroke; two examples being “the first line is close to representative stroke pattern No.
1
or representative stroke pattern No.
2
” and “the beginning of the second line is close to and below the beginning of the first line.”
In the prior-art device described in Reference
1
, the patterns of stroke units are quantized without alteration and therefore include information on stroke shape and position. Patterns of differing position but of similar shape are therefore prepared as different stroke unit patterns, and this prior-art method therefore necessitates a voluminous stroke code book, a huge amount of memory for maintaining the stroke code book, and numerous calculations for finding inter-stroke distances.
In the method described in Reference
2
, there is the problem that a recognition dictionary must be prepared relating to each of different characteristics for every character. The production of these recognition dictionaries requires operator input for each individual character, and therefore necessitates a great deal of labor. Finally, there is the problem that control of the recognition process requires a complicated process of solving fuzzy propositions for every character, and this type of process is difficult to realize a high-speed and repetitive process by hardware.
SUMMARY OF THE INVENTION
The object of the present invention is to provide an on-line character recognition method and device that allow a great reduction in the amount of required memory, the number of computations, and number of steps required for preparing a recognition dictionary.
To achieve the above-described objects, the present invention provides an on-line character recognition method that includes the steps of:
inputting the stroke patterns of an inputted character;
finding a stroke shape pattern that indicates stroke shape and a stroke position pattern that indicates stroke position or size for each inputted stroke pattern;
finding inter-stroke shape distance, which is the distance between each stroke shape pattern of the inputted character and a stroke shape pattern obtained in advance with respect to the stroke pattern of a reference character, and inter-stroke position distance, which is the distance between each stroke position pattern of the inputted character and a stroke position pattern obtained in advance with respect to the stroke pattern of a reference character; and
finding inter-stroke distance between strokes of each target character and strokes of the inputted character based on the inter-stroke shape distances and the inter-stroke position distances, and taking as the recognition result for the inputted character the target character corresponding to the smallest inter-stroke distance.
In addition, the present invention provides an on-line character recognition device that includes:
pattern detection means for inputting stroke patterns of an inputted character, and finding from each stroke pattern a stroke shape pattern indicating stroke shape and a stroke position pattern indicating stroke position or size;
inter-stroke shape distance calculation means for finding inter-stroke shape distance, which is the distance between each of the stroke shape patterns and stroke shape patterns obtained in advance with respect to stroke patterns of reference characters;
inter-stroke position distance calculation means for finding inter-stroke position distance, which is the distance between each of the stroke position patterns and stroke position patterns obtained in advance with respect to stroke patterns of reference characters; and
matching means for finding inter-stroke distance between strokes of each target character and the strokes of the inputted character based on the inter-stroke shape distance and the inter-stroke position distance and taking as the recognition result for the inputted character the target character having the smallest inter-stroke distance.
In an on-line character recognition that takes strokes as units, strokes can be represented by dividing between patterns corresponding to shape and patterns corresponding to the position of strokes, and information common to various characters can therefore be compactly organized by preparing code books for each pattern.
This arrangement allows processing by simple distance calculations, thereby allowing a reduction in the number of calculations and the amount of memory, and moreover, allowing a simplification of the production of a dictionary.
REFERENCES:
patent: 4607386 (1986-08-01), Morita et al.
patent: 4718102 (1988-01-01), Crane et al.
patent: 5588074 (1996-12-01), Sugiyama
patent: 5745599 (1998-04-01), Uchiyama et al.
patent: 55-138171 (1980-10-01), None
patent: 60-68484 (1985-04-01), None
patent: 60-237580 (1985-11-01), None
patent: 62-164183 (1987-07-01), None
patent: 62-269089 (1987-10-01), None
patent: 1-136288 (1989-05-01), None
patent: 1-56889 (1989-06-01), None
patent: 4-336686 (1992-11-01), None
Papers of the Institute of Electronics, Information and Communication Engieneers, Feb. 1991, D-II, vol. J74-D-II, No. 2, pp. 166-174, “Application of Fuzzy Logic in an On-line Handwritten Character Recognition Method”.
Ikebata Yoshikazu
Nakashima Yutaka
Yoshida Kazunaga
Couso Jose L.
Do Anh Hong
NEC Corporation
Sughrue Mion Zinn Macpeak & Seas, PLLC
LandOfFree
On line-character recognition method and device 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 method and device, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and On line-character recognition method and device will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-2484700