JPH0330191B2 - - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION (Technical Field) The present invention relates to a method for reading handwritten documents at high speed and with high accuracy.

(Background technology) As a method for reading handwritten documents, the presence or absence of a specification in a writing frame that specifies the character type of the written character is detected, and the characters in the column where the specification is detected are read only in a dictionary of the specified character type. A method has been proposed that refers to However, with this method,
If a character type is not specified, all dictionaries must be referenced, resulting in a slow processing speed.

(Objective and Summary of the Invention) An object of the present invention is to improve the above-mentioned drawbacks of the conventional technology and provide a high-speed and highly accurate handwritten document reading method. This is taken as the complexity of the character, and if a character type is not specified, the character is identified by selecting a dictionary of the character type that includes the character based on the complexity.

(Embodiment of the Invention) FIG. 1 is a block diagram showing an embodiment of a handwritten Japanese document reading method according to the present invention. In the figure, 1
is a photoelectric conversion unit, 2 is a pattern register, 3 is a feature extraction unit, 4 is a character dose detection unit, 5 is a character type detection unit,
6 is an identification unit, 7 is a character name output, 8 is a hiragana dictionary memory, 9 is a katakana dictionary memory, 10 is an alphanumeric symbol dictionary memory, and 11 is a kanji dictionary memory.

Fig. 2 is a diagram showing an example of a form used in this example, where 21 is a form, 22 is a character type specification entry frame, 23 is a kanji specification field, 24 is a hiragana specification field, and 25 is a katakana specification field. 26 is an alphanumeric symbol designation field, 27 is a character entry frame, 28 is a character type designation entry line, and 29 is a character entry line. For example, entries are made as in the form entry example shown in FIG. 3.

The operation of the present invention will be explained below using this example of the form.

First, enter line 28 for character type specification in form 21 in Figure 2.
The presence or absence of character type designation corresponding to each character in the character entry line 29 is detected for the line , and the presence or absence of character type designation is sent to the identification unit 6. In this operation, the photoelectric conversion unit 1 performs photoelectric conversion on the character type designation entry line 28 to convert it into a binary quantized electric signal, cuts out an area for one character, and stores it in the pattern register 2. The character type designation detection unit 5 divides the pattern register 2 into four areas corresponding to the character type designation entry frame 22, counts the number of black dots in each area (character line parts are black dots), and calculates the threshold value. The comparison is made to detect whether or not each character type is designated, and the character type, that is, kanji,
The presence or absence of designation of hiragana, katakana, symbol, etc. is stored in the character type designation memory in the character type detection unit 5. Through the above operations, it is detected whether or not a character type has been specified corresponding to each character in the character entry line 29. Next, the character entry line 29 in FIG. 2 is read. The operation is such that the photoelectric conversion unit 1 performs photoelectric conversion on the character entry line 29, converting it into a binary quantized electric signal,
An area corresponding to one character is cut out and stored in the pattern register 2. The feature extraction section 3 extracts various features from the character patterns in the pattern register 2 and sends the features to the identification section 6.

At the same time, the pattern register 2 in the character dose detection unit 4
The character complexity level D is determined by detecting the character radiation dose from the character pattern within and normalizing it by the character size. The complexity is expressed by the following equation.

D=A×K/WL×(PB+PR) However, K is a constant to convert D into an integer, A is the total number of black dots in the character frame, PB is the size of the circumscribed frame of the character in the height direction, and similarly PR indicates the size in the width direction. WL is the line width of the character and is calculated using the following formula.

WL=A/A-Q However, Q represents the number of points in which all four points are black points when observing all points within the character frame using a 2×2 window.

When the complexity level D of the character is detected, the complexity level D is sent to the character type designation detection section 5. The character type designation detection unit 5 sequentially refers to the character type designation memory and sends the character type designation to the identification unit 6. If the character type designation cannot be detected in the character type designation area as shown at 22 in FIG. The following conditions are determined, the character type is determined, and the character type is sent to the identification unit 6.

D<a See all dictionaries.

D≧a The character type is assumed to be a kanji, and a kanji dictionary is referred to.

However, in this example, a=10 and K=5.

The identification unit 6 compares the features sent from the feature extraction unit 3 with a dictionary, and finally outputs a one-character color name to the character name output 7.

The dictionary memories used in the identification unit 6 include a hiragana dictionary memory 8, a katakana dictionary memory 9, an alphanumeric symbol dictionary memory 10, and a kanji dictionary memory 11.
Four types are prepared, and the features sent from the feature extraction section 3 are compared with the dictionary using the dictionary memory of the character types in which the character types corresponding to each character have been designated in advance.

(Effects of the Invention) As described above in detail, the present invention detects the character type designation in advance, and for characters without the above designation, the character type is selected by detecting the character dose of the character. Since characters are identified using a suitable dictionary, reading can be performed at high speed and with high precision, and therefore, handwritten Japanese documents can be read at high speed and with high precision.

[Brief explanation of drawings]

FIG. 1 is a block diagram showing an embodiment of the handwritten document reading method according to the present invention, FIG. 2 is a diagram showing an example of a form used in the embodiment of the present invention, and FIG. 3 is a diagram showing an example of filling out the form. It is. 1...Photoelectric conversion unit, 2...Pattern register, 3
...Feature extraction unit, 4...Character dose detection unit, 5...
Character type designation detection unit, 6...Identification unit, 7...Character name output, 8...Hiragana dictionary memory, 9...Katakana dictionary memory, 10...Alphabet/numeric symbol dictionary memory, 11...Kanji dictionary memory, 21... …Form, 2
2...Character type specification entry box, 23...Hiragana specification field, 24...Katakana specification field, 25...Alphabet, numeric symbol specification field, 26...Kanji specification field, 27...Character entry box, 28...Character type specification Entry line, 29...Character entry line.

Claims (1)

[Scope of Claims] 1. A handwritten Japanese document has a character frame in which characters are written, and a character type specification area provided near the character frame to specify a character type, and a dictionary of the character type specified in the character type specification area. In a handwritten document reading method that recognizes handwritten characters written in a character frame, when the character type is not specified, the character dose of the handwritten character is determined as the complexity of the character, and a dictionary of character types corresponding to the complexity is obtained. 1. A handwritten document reading method comprising: selecting a dictionary; and recognizing handwritten characters using the selected dictionary. 2. The handwritten document reading method according to claim 1, wherein the dictionary is created for hiragana, katakana, alphanumeric characters, and kanji, respectively.