JPS60118985A - Character recognizing system - Google Patents

Abstract

PURPOSE:To switch automatically processing system of character recognition by comparing a slope of a character in 10-100 character groups read at the beginning of an original with a slope of a reference print character so as to discriminate it as manuscript on the original when the former slope is larger. CONSTITUTION:A character recognition switching control section 4 switching the processing system in a character recognition section 3 and an extracting section 5 extracting the line width (slope) of each character in a prescribed character group reat at the beginning in the original are provided. Then a line width signal (slope signal) transmitted sequentially from the extracting section 5 and a preset reference signal Sr are compared by a system controller and the number of characters when the relation of S>Sr exists is counted by a count means 7, the number is compared and discriminated with the reference character number so as to discriminate whether the character in the original is a print character or a manuscript.

Description

DETAILED DESCRIPTION OF THE INVENTION The present invention relates to a character type identification method for identifying whether characters (including patterns such as symbols) targeted for pattern recognition in OCR are printed or handwritten.

In general, handwritten characters are more deformed than printed characters, making the processing during character recognition in OCR more complicated. However, in order for OCR to be able to recognize both printed characters and handwritten characters, the processing time of OCR is determined by the processing time for recognizing handwritten characters, which slows down the processing speed. There is a problem with this.

Therefore, in order to reduce the processing time for character recognition using printed characters, conventionally, two systems were provided for character recognition processing in OCR so that printed characters and handwritten characters could be processed separately. Then, the operator selects a switch in advance to instruct the OCR controller whether the input characters to be recognized are printed or handwritten, and the controller switches the character recognition processing system as appropriate to recognize the input characters. I'm trying to get them to recognize it. However, using such a switch selection means makes the operation complicated.

IJke The present invention was made in consideration of the above points, and the OCR automatically determines whether the input characters are printed or handwritten, and then automatically determines whether the input characters are printed or handwritten. The present invention provides a character type identification method that allows automatic switching of character recognition processing systems.

room An embodiment of the present invention will be described in detail below.

The character type identification method according to the present invention takes into consideration the fact that handwritten characters generally have a larger line width than printed characters, and uses the first 10 to 100 Measure the line width of each character in the character group.

Compare the line width of each requested character with the average or maximum line width of standard printed characters, and if there are more than a predetermined percentage of characters whose requested line width is larger than the standard line width, The document is determined to be handwritten, and if not, it is determined to be printed.

Further, in the character type identification method according to the present invention.

Focusing on the fact that handwritten characters generally have a larger slope than printed characters, we calculated the slope of each character in the first 10 to 100 characters read in the manuscript. The slope of each character is compared with the average or maximum slope of the standard printed characters, and if the number of characters whose slope is greater than the standard slope is more than a predetermined percentage, The document is determined to be handwritten, and if not, it is determined to be printed.

FIG. 1 shows an example of a configuration in which the character type identification method according to the present invention is specifically implemented in OCR, in which a scanner 1 optically scans a document surface under the control of a system controller consisting of a microcomputer or the like. The image information is read pixel by pixel, the character cutting section 2 cuts out each character according to the read image information, and the character recognition section 3 separates printed characters and handwritten characters.
In an OCR that performs character recognition by extracting the characteristics of each character cut out into systems, the character recognition switching control section 4 specifically switches between each processing system in the character recognition section 3; The line width (or slope) of each character in a predetermined group of characters first read in the manuscript is one of the features extracted from each character in the character recognition unit 3.
A line width extraction unit (or slope extraction unit) 5 is provided to extract the line width extraction unit (or slope extraction unit) 5, and on the system controller side, the line width signal (or slope signal) S sequentially sent from the line width extraction unit 5 and a preset reference line are Width signal (or reference slope signal) Sr
Comparison means 6 for determining the magnitude relationship between the two
and a counting means 7 which counts and stores the number of characters for which S>Sr according to the comparison result, and compares and judges the counted number of characters with a preset standard number of characters and determines whether the characters in the manuscript are correct. A character type determining means 8 determines whether the character is printed or handwritten and gives a switching command to the character recognition switching control section 4.

However, in the device configured in this way, each stroke information for the first 10 characters in the document read by the scanner 1 and cut out by the single character cutting unit 2 is sequentially sent to the line width extraction unit 5. Then, the line width of each character is extracted. in particular.

For example, line width = total black pixels/perimeter length.

Next, the comparing means 6 successively compares each set line width S with the reference line width Sr, and the counting means 7 counts the number of high-level output signals when 'S>Sr. Next, based on the total count value, the character type determining means 8 determines whether there are, for example, five or more characters whose line width S is equal to or greater than the reference line width Sr, and if there are five or more characters, the number of characters in the manuscript is It is assumed that the characters are handwritten, and if not, they are printed, and the character recognition switching control unit 4 is given switching commands to that effect to cause the character recognition unit 3 to switch modes as appropriate. . This allows the character recognition unit 3 to perform optimal character recognition processing depending on whether the characters are printed or handwritten. Therefore, especially when recognizing printed characters, the dictionary search in the character recognition section 3 is limited to printed characters only, and the processing time is effectively shortened. It should be noted that character recognition for the first 10 characters in the manuscript will be processed in both print and handwritten character modes. Further, FIG. 2 shows the procedure of operation of each part.

In the character type identification method according to the present invention, OCR
Automatically determines whether input characters are printed or handwritten, and automatically switches the character recognition processing system according to the determined type of character. This method has the excellent advantage of being able to quickly perform optimal character recognition processing according to the character type without requiring any complicated operations.

[Brief explanation of the drawing]

FIG. 1 is a block diagram showing a configuration example for concretely implementing the character type identification method according to the present invention, and FIG. 2 is a flow chart of the operation of each part in the same configuration example. ■...Scanner 2...Character cutting section 3...Character recognition section 4...Character recognition switching control section 5...Line width extraction section 6...Comparison means 7...Counting means 8・・・
・Character type determination method applicant's representative Kiyoshi Torii 1 Figure 7 8 Figure 2

Claims (1)

[Claims] 1. Means for determining the line width of each character in a group of characters extracted from a manuscript, and a reference line width that is set in advance according to the type of printed character to determine the line width of each character. If there is a predetermined percentage of characters whose line width is larger than the standard line width based on the comparison result, it is determined that the manuscript is handwritten, and if not, it is determined that the manuscript is written as printed characters. A character type identification method comprising means for determining that the character type is 2. A means for determining the slope of each character in a group of characters extracted from a manuscript, a means for comparing the determined slope of each character with a reference slope preset according to the type of printed characters, and When there are more than a predetermined percentage of characters whose slope determined from the comparison result is larger than the standard slope, the manuscript is determined to be handwritten, and if not, it is determined to be printed. A character type identification method consisting of means.