JPH01259476A - Character reader - Google Patents

Abstract

PURPOSE:To improve the processing efficiency of a character reader by sectioning a field into the field of a handwriting character and the field of a printing type character and recognizing the character strings of the respective fields by being collected. CONSTITUTION:According to the instruction of an instructing means 21, a control part 20 of a recognizing part 22 investigates whether a character pattern string to be stored in a picture memory 5 is read from the handwriting character field or read from the printing type character field based on the format information of a document given in advance. Whether there is the character pattern string to be read from the plural handwriting character fields or not and there is the character pattern string to be read from the plural printing type character fields or not are investigated. Then, the control part 20 divides the field into the field of the handwriting character and the field of the printing type character based on the instruction of the instructing means 21 and the character strings of the respective fields are recognized all at once. Accordingly, an idle time generated in a recognizing part per field is totally shortened. Thus, the processing efficiency of the character reader can be improved.

Description

[Detailed Description of the Invention] [Summary] It is equipped with a plurality of recognition units that perform character matching and recognition. Regarding character reading devices that increase processing efficiency, the aim is to reduce play time that occurs in the recognition unit and increase character recognition processing efficiency. After distinguishing and recognizing the areas in which the printed characters are written, a recognition unit reads out a plurality of character pattern strings stored in the image memory and differentiated for each area, and recognizes each character pattern in the character pattern string. A character reading device comprising a plurality of characters, the recognition unit having an instruction to continuously read out a plurality of character pattern strings read from a plurality of areas on a paper on which handwritten characters are written, and a paper on which printed characters are written. The character pattern string read from the area on the paper on which the handwritten characters are written on the image memory is provided with an instruction means for instructing to successively read out a plurality of character pattern strings read from the plurality of areas on the image memory. If there are multiple character pattern strings, character recognition is performed by sequentially reading out multiple character pattern strings, and multiple character pattern strings read from the area on the paper where the printed characters are written on the image memory are read out continuously. If it exists, the configuration is such that a plurality of character pattern strings are read out continuously and character recognition is performed.

[Industrial application field]

The present invention relates to a character reading device equipped with a plurality of recognition units that collate and recognize characters, and in particular, a field for writing handwritten characters (a character writing area provided on the paper) and a field for writing printed characters are provided on the paper. If there are multiple in one line,
The present invention relates to a character reading device that improves recognition processing speed.

In recent years, with the development of an information-oriented society, the amount of data input to computer systems has increased, and higher-speed data input devices have been required. For this reason, character reading devices also need to be able to read characters at higher speeds, by cutting out characters one by one from a character string read from multiple fields in one line of paper, and comparing them with a standard pattern, for example. Character reading devices are used that are equipped with a plurality of recognition units that perform character recognition and perform high-speed processing.

By the way, it is necessary for these plurality of recognition units to operate efficiently at all times.

[Conventional technology]

FIG. 5 is a block diagram explaining the conventional technique, and FIG. 6 is a diagram explaining the operation of FIG. 5.

Light is projected from a light source 1 onto, for example, a form 2, and the reflected light is focused by a lens 3 to form an image on a photoelectric conversion binarization circuit 4. The photoelectric conversion binarization circuit 4 converts the reflected light into an electrical signal using, for example, a CCD, and as the form 2 moves in the direction of arrow A, a field as shown in FIG. 6(a) on the form 2 is displayed. Read the written characters. The data converted into electrical signals is then binarized and stored in the image memory 5 as a character pattern string.

In form 2, as shown in FIG. 6(a), for example, field (1), field (2), and field (3) exist in one line of the form, and field (1) and field (
Assuming that field (3) is a field in which handwritten characters are entered and field (2) is a field in which printed characters are entered, a character pattern string is stored in the image memory 5 for each field.

The control unit 10 of the recognition unit 6 has a field [11f31
Field information of form 2 is given that indicates that field (2) is a field in which handwritten characters are entered, and field (2) is a field in which printed characters are entered. , because the processing up to character recognition is different, the control unit 10 performs the following for each field (11 (21f3)) stored in the image memory 5:
Each character pattern string is read out and input to the character cutting section 11.

That is, first, the character patterns ■ to ■ read from field (1) of the form 2 and stored in the image memory 5 are successively displayed.
These character patterns ■ to ■ are input to the character cutting section 11.

Based on the control of the control unit 10, the character cutting unit 1 continuously cuts out characters from the character patterns ■ to ■ inputted from the image memory 5, as shown in FIG. 6(b). When the character pattern ■ is cut out, it is sent to the switching unit 12.

The switching unit 12 sends this character pattern ■ to the feature extraction unit 17 of the recognition unit 7, as shown in FIG. 6 (C1).

This is because the control unit 10 controls character extraction after the next character pattern.
This is because 5 cannot be controlled.

The feature extraction section 17 extracts the features of the character pattern ■ under the control of the control section 16 and sends it to the matching section 18 . Collation part 1
8 reads the features of the standard pattern from the dictionary 8 and compares them. Then, a plurality of candidate characters in the dictionary 8 are selected in descending order of feature dissimilarity and sent to the determination unit 19.

The determining unit 19 finally selects candidate characters based on the format information of the form 2 given in advance, such as whether the field (11 is a number or an alphabetic character), and the editing unit 9
Send to.

The character pattern ■ cut out next by the character cutting section 11 is
The switching section 12 sends the result to the feature extraction section 17, and the determination result is sent to the editing section 9 in the same manner as described above.

The similarly extracted character pattern ■ is sent to the feature extraction unit 13 via the switching unit 12 because the character extraction operation of the character extraction unit 11 has now completed up to the character pattern ■, as shown in FIG. 6 tbl. The features of the character pattern ■ extracted by the feature extraction section 13 are compared with the features of the standard pattern in the dictionary 8 by the matching section 14. Then, the candidate characters in the dictionary 8 are sent to the plurality determination unit 15 in order of decreasing degree of difference, and the candidate characters are finally selected based on the format information of the form 2 given in advance in the same manner as described above. Sent out.

The editing section 9 edits the characters of the recognition results sent from the recognition sections 6 and 7 in the order of ■ to ■ based on the information given from the control section 10.

Regarding field (2) shown in FIG. 6 (al), since the recognition is of printed characters, the control programs of the control units 10 and 16 are different, and the extracted features in the feature extraction units 13 and 17 and the contents of the dictionary 8 are also different. , as above, Fig. 6(C)
As shown in FIG. 6, the character patterns ■ and ■ are recognized by the recognition unit 7, and as shown in FIG. 6(b), the character patterns ■ and ■ are recognized by the recognition unit 6.
and [phase] are recognized and sent to the editing section 9. and,
The editing section 9 edits these characters in the order of ■ to [phase].

The case of field (3) shown in FIG. 6(al) is exactly the same as the case of field (11), and as shown in FIG. 6(b), the character pattern @ and [phase] are recognized by the recognition unit 6, and 6
As shown in FIG.

[Problem to be solved by the invention]

As mentioned above, conventionally, it is not possible to proceed to the recognition process of the next field (2) until the recognition process of field (1) is completed, and similarly, after the recognition process of field (2) is completed, The process now moves on to recognition processing for field (3).

Therefore, the recognition unit that completes the recognition process quickly for each field becomes idle, and when processing a form in which fields of handwritten characters and fields of printed characters coexist in one line,
- There is a problem that the processing efficiency decreases as the number of fields in a row increases.

In view of these problems, the present invention combines fields (1) and (3) of handwritten characters and continuously creates an image memo U5.
The process from character segmentation to recognition is completed by reading out the printed character field (2), and the process from character segmentation to recognition is completed by reading out the printed character field (2). The purpose is to increase processing efficiency.

[Means to solve the problem]

FIG. 1 is a block diagram of the principle of the present invention.

The same reference numerals as in FIG. 5 indicate the same functions. Recognition unit 22
The control unit 20 controls the image memory 5 based on the format information of the form given in advance based on the instruction from the instruction means 21.
Whether the string of character patterns stored in the Check if there is a string of character patterns read from the print character field.

If there are character pattern strings read from a plurality of handwritten character fields, the plurality of character pattern strings are collectively read out continuously from the image memory 5 and input into the character cutting section 11.

The control unit 20 sends each cut-out character pattern to the recognition unit 7 via the switching unit 12 until the character cutting of a series of character pattern strings is completed, and then sends it alternately to the feature extraction unit 13 to determine the character pattern. The matching unit 14 extracts the features of
The candidate characters are compared with the characteristics of the dictionary 8, and the candidate characters are sent to the determination section 15, which selects the candidate characters based on the format information. Then, the characters selected from the determining section 15 and the recognizing section 7 are sent to the editing section 9 to edit the character string for each field.

The same applies if there is a character pattern sequence read from a printed character field.

[Effect]

With the above configuration, the control section 20 separates the field of handwritten characters and the field of printed characters based on the instruction from the instruction means 21, and recognizes the character strings in each field at once. The play time that occurs in the recognition unit 22 or 7 for each field is collectively shortened.

Therefore, the processing efficiency of the character reading device can be improved.

〔Example〕

FIG. 2 is a block diagram of a circuit showing an embodiment of the present invention.
FIG. 3 is a flowchart for explaining the operation of FIG. 2, and FIG. 4 is a diagram for explaining the operation of FIG. 2.

The same reference numerals as in FIG. 5 indicate the same functions. Recognition unit 22
Based on the instructions from the instruction program 23, the control unit 20 inputs the format information of the form given in advance, as shown in FIG. [Based on this, a handwritten character field and a printed character field are distinguished.

That is, the character pattern string stored in the image memory 5 is recognized and classified as to whether it is read from a handwritten character field or a printed character field. Then, check to see if there is a handwritten character field.

If there is no handwritten character field, the routine moves on to check whether there is a printed character field, and if there is a handwritten character field, it is checked whether there are multiple handwritten character fields.

If there is not a plurality of character pattern strings, the character pattern strings of the relevant field are read out, and if there are a plurality of character pattern strings of the field, the character pattern strings of the plurality of fields are collectively read out one after another and the characters are recognized.

That is, assuming that fields (1) and (3) shown in FIG. up to and input into the character cutting section 11. The character cutting unit 11 continuously cuts out characters as shown in FIG. 4(b) under the control of the control unit 20, and after cutting out characters ■ to ■, respectively, it continues cutting out characters ■ to ■, respectively. .

The control unit 20 sends the characters ■, ■, and ■, which are cut out during character cutting, for example, to the recognition unit 7 through the switching unit 12 for recognition, and then alternately passes through the switching unit 12 and converts the characters ■■ 0 ■
is sent to the feature extraction unit 13 to extract the features, the matching unit 14 reads out the features of the standard pattern from the dictionary 8 and matches them, and sends candidate characters to the determination unit 15. The determining unit 15 selects candidate characters based on the format information and sends them to the editing unit 9.

Further, the control unit 20 sends the character ■00 [phase] to the recognition unit 7 for recognition. The editing section 9 edits the characters sent out by the judgment section 15 and the judgment section of the recognition section 7 for each field based on the format information, and sends them out as a character string for each field.

As shown in FIG. 3, the control section 20 then cycles through four cycles until there is a printed character field, and if there is no printed character field, the operation is terminated, and if there is a printed character field, it is checked whether there is more than one. If there is not a plurality of character pattern strings, the character pattern strings of the relevant field are read out, and if there are a plurality of character pattern strings of the field, the character pattern strings of the plurality of fields are collectively read out one after another and the characters are recognized.

That is, assuming that the field (2) shown in FIG. Under the control of the unit 20, characters are continuously cut out as shown in FIG. 4(b), and the characters
Cut out each part from to 0.

The control unit 20 sends the character ■, for example, which is cut out during character extraction, to the recognition unit 7 through the switching unit 12 for recognition, and then alternately passes through the switching unit 12 and sends the character ■0 to the feature extraction unit. 13 to extract the features, and the matching unit 14 reads out the features of the standard pattern from the dictionary 8 and matches them.
The candidate characters are sent to the determination unit 15. The determining unit 15 selects candidate characters based on the format information and sends them to the editing unit 9.

Further, the control unit 20 sends the character ■ to the recognition unit 7 for recognition. The editing section 9 edits the characters sent out by the judgment section 15 and the judgment section of the recognition section 7 for each field based on the format information, and sends them out as a character string for each field.

〔Effect of the invention〕

As explained above, when characters are read from a form or the like in which fields of handwritten characters and fields of printed characters coexist, the present invention separates the fields into the field of handwritten characters and the field of printed characters, and reads the characters in each field. Since the columns are recognized all at once, the idle time of the plurality of recognition units is reduced, and the processing efficiency of the character reading device can be improved.

[Brief explanation of the drawing]

FIG. 1 is a principle block diagram of the present invention, FIG. 2 is a block diagram of a circuit showing an embodiment of the present invention, FIG. 3 is a flowchart explaining the operation of FIG. 2, and FIG.
5 is a block diagram illustrating a conventional technique, and FIG. 6 is a diagram illustrating the operation shown in FIG. 5. In the figure, ■ is a light source, 2 is a form, 3 is a lens, 4 is a photoelectric conversion binarization circuit, 5 is an image memory, 6, 7.22 is a recognition unit, 8 is a dictionary,
9 is the editing section, 10, 16.20 is the control section, 11 is the character cutting section, 12
is the switching section, 13.17 is the feature extraction section, 14, 18
15.19 is a determination unit, 21 is an instruction means,
23 is an instruction program. Is the 20th invitation official Sun Moon T Ro Florape? -Toro 3
Mouth (Part 1 Plan 2 Diagram 0)
Henguchi・2)

Claims (1)

[Scope of Claims] Based on the format information of the paper given in advance, the areas on the paper where handwritten characters and printed characters are written are distinguished and recognized, and then each area is distinguished and stored in the image memory. (5) A character reading device comprising a plurality of recognition sections (22) and (7) that read out a plurality of character pattern strings stored in the character pattern string and recognize each character pattern of the character pattern string, respectively, the recognition section (22) an instruction to successively read out a plurality of character pattern strings read from a plurality of areas on the paper on which handwritten characters are written, and a plurality of character pattern strings read from a plurality of areas on the paper on which printed characters are written; an instruction means (21) for instructing to read out consecutively a plurality of character pattern strings written on the image memory (5); If there are multiple character pattern sequences, character recognition is performed by sequentially reading out the multiple character pattern sequences, and the image memory (5)
) If there are a plurality of character pattern strings read from the area on the paper on which printed characters are written, character recognition is performed by sequentially reading out the plurality of character pattern strings together. Character reading device.