JPH041882A - Character recognition and device for the same - Google Patents

Abstract

PURPOSE:To improve character recognition speed by retrieving a dictionary memory based on a character identification number registered in a character registering memory through a character decision circuit, selecting the character satisfying decision items and specifying the type of character. CONSTITUTION:A character decision circuit 10 judging whether or not the contents of a buffer memory 8 satisfies the decision items of a dictionary memory 9 and outputs the recognition number of a character satisfying all the decision items. In the recognition of an ordinary recognition area 13 of a mark sheet 1, character identification numbers are sequentially picked out of a character registering memory 12, inputted to the character decision circuit 10 and the decision items of the dictionary memory 9 corresponding to the character are read. When the character satisfies the decision items or such a character is not found, a character in which the cumulative deviation of the decision items is minimum is selected to decide the character. Thus, time required for retrieving the dictionary can be decreased without increasing the capacity of the dictionary memory, so that recognition speed can be improved.

Description

DETAILED DESCRIPTION OF THE INVENTION [Field of Industrial Application] The present invention relates to a character recognition method and apparatus, and more particularly to a character recognition method and apparatus for recognizing characters handwritten on a mark sheet.

[Conventional technology]

Conventionally, this type of character recognition device has an extraction circuit that extracts a single character pattern from mark sheet screen information, and a
There is a feature extraction circuit that extracts partial features necessary for character recognition from character patterns, a feature counting circuit that digitizes the degree of each partial feature, and a feature counting circuit that digitizes the degree of each partial feature. It has a dictionary memory that stores the character type to which it belongs, the partial feature number to be focused on as each judgment item, its reference value, and tolerance value, and the character judgment circuit searches the judgment items of all the fonts written in the dictionary memory. , compare the counted value of the partial feature output from the feature counting circuit with the reference value, determine whether the judgment item is satisfied based on whether it is within the tolerance value, select the font that satisfies all the judgment items, and select the font that satisfies all the judgment items. The character type to which the font belongs is output as a recognition result.

[Problem to be solved by the invention]

In the judgment logic of the conventional character recognition device described above, if there is a large deviation between the characteristics of the sample font and the written font, judgment errors are likely to occur, so in order to increase the correct answer rate, the sample font is However, this increases the capacity of the dictionary memory and takes processing time.

[Means to solve the problem]

In the character recognition method of the present invention, a training area is provided in a mark sheet, the cutting circuit recognizes the training area, the character type to be written in the training area is determined based on the position of the writing frame, and When performing recognition processing on the written characters, the character judgment circuit selects a font identification number that satisfies all the judgment items written in the dictionary memory, stores it in the font registration memory, and stores it in the normal area of the mark sheet after training. In recognition, the character determination circuit selects a font that satisfies determination items by searching the dictionary memory based on the font identification number registered in the font registration memory and identifies the character type. It is characterized by

The character recognition device of the present invention also includes an extraction circuit that extracts pattern information for one character from mark sheet screen information;
a feature extraction circuit that extracts partial features necessary for character determination from a single character pattern; a feature counting circuit that digitizes the degree of each of the partial features; and a feature counting circuit that digitizes the degree of each of the partial features; A dictionary memory that stores the character type to which it belongs, the partial feature number to be focused on as each judgment item, its reference value, and allowable deviation value, and the output of the feature counting circuit is collated according to the judgment items of this dictionary memory to determine all judgment items. a character recognition circuit that selects a font that satisfies the above, and a training area pressure detection means that detects a training area in which a character type to be written is fixed based on a writing frame position provided in the mark sheet;
and a font registration memory, and the processing of the character judgment circuit for the characters written in the training area searches for all the fonts of the corresponding character type described in the dictionary memory, and satisfies the judgment items. selecting a font identification number and storing it in the font registration memory;
The character determination circuit processes the normal area of the mark sheet after the training process by searching the dictionary memory based on the font identification number registered in the font registration memory. A special feature characterized by selecting a font that satisfies the determination items and specifying the character type [Embodiment] Next, the present invention will be described with reference to the drawings.

FIG. 1 is a block diagram showing an embodiment of the present invention, FIG. 2 is a plan view showing an example of a mark sheet used in this embodiment,
FIG. 3 is a flowchart illustrating an example of the determination logic in the training mode in this embodiment, and FIG. 4 is a flowchart illustrating an example of determination logic in the normal recognition mode in this embodiment.

The mark sheet 1 is converted into an electrical signal using the optical reading section 2, and written into the input image data buffer memory 3 as two-dimensional pattern information, and the cutting circuit 4 detects the cutting position mark 5 printed on the mark sheet 1 and reads one character. Character patterns are cut out one by one and transferred to the feature extraction circuit 6.

The feature extraction circuit 6 detects partial features necessary for character determination from the character pattern, and sends the information to the feature counting circuit 7.

The feature counting circuit 7 converts the degree of each partial feature into a numerical value, and stores the value for one character in the buffer memory 8.

On the other hand, the dictionary memory 9 stores in advance partial feature numbers, their reference values, and allowable deviation values as judgment items for each sample font.

The character determination circuit 10 determines whether the content of the buffer memory 8 satisfies the determination items in the dictionary memory 9, and outputs the recognition number of the font that satisfies all the determination items.

The operation up to this point is the same as the conventional example, but in this example, the training area designation mark 11 on the mark sheet 1
is detected by the extraction circuit 4, and the character determination circuit 10 is notified that it is the training mode.

During the training mode, the font identification numbers output from the character determination circuit 10 are classified into each character type and stored in the font registration memory 12.

In the recognition of the normal recognition area 13 of the mark sheet 1, the font identification numbers are sequentially retrieved from the font registration memory 12 and input to the character judgment circuit 10, and the judgment items in the dictionary memory 9 corresponding to the fonts are read out and the judgment items are read out. If the result is satisfied, or if such a font is not found, a font with the minimum cumulative M deviation for each determination item is selected and character determination is performed.

First, font determination during training mode is processed according to the flow shown in FIG. That is, the font identification number p is obtained from the character type g output from the character judgment circuit 10 (step 31), the start address n of the judgment item for the font p is obtained from the dictionary table of contents (step 32), and then the font identification number p is obtained from the dictionary table of contents (step 32). Step 33) Read partial feature number a, feature reference value, and allowable deviation value C from address n, buffer memory 8
The count value d corresponding to the partial feature number a is read from (step 34). Then, when the difference between the characteristic reference value and the count value d is smaller than the allowable deviation value C, the address n is updated (steps 35 and 36). Next, it is determined whether it is the end of the judgment item for font p (step 37), and if it is the end, font p is written into the font registration memory 12 (step 38), but if it is not the end in step 37, steps 33 to 37 are stored as addresses. Repeat while updating n. After the writing is completed in step 38, if it is the end of the font of character type g, the character determination ends, and if it is not the end of the font, the character identification number p
is updated and the flow from step 32 is repeated (step 39).

Next, character determination in the normal recognition mode is processed according to the flow shown in FIG. That is, the character determination circuit 10 reads the font identification number p from the address k of the character registration memory 12 (step 41), and obtains the start address n of the characteristic description of the font p from the dictionary table of contents (step 42). Then step 33 as described above. ~ Perform the same processing as in 36 (
Step 43. ~46), it is determined whether or not the end of the feature description of the font p is reached (Step 47), and if it is the end, the font p
The result is output in step 48) and the process ends. If the process does not end here, the process from step 43 is repeated.

In this way, in this embodiment, character determination is performed for the written fonts, reflecting the tendency of the fonts.

〔Effect of the invention〕

As explained above, the present invention improves the correct answer rate in character identification by providing a training mode in which the person filling out the mark sheet writes in their own handwriting, and the time required for dictionary searches can be saved without increasing the capacity of the dictionary memory. It has the effect of improving recognition speed.

[Brief explanation of drawings]

FIG. 1 is a block diagram showing an embodiment of the present invention, FIG. 2 is a plan view showing an example of a mark sheet used in this embodiment,
FIG. 3 is a flowchart illustrating an example of the determination logic in the training mode in this embodiment, and FIG. 4 is a flowchart illustrating an example of determination logic in the normal recognition mode in this embodiment. 1... Mark sheet, 2... Optical reading section, 3...
Input image data buffer memory, 4... cutout circuit,
5... Cutting position mark, 6... Feature extraction circuit,
7... Feature counting circuit, 8... Buffer memory, 9...
...Dictionary memory, 10...Character judgment circuit, 11...
Training area designation mark, 12... Memory for font registration, 13... Normal recognition area, 14... Training area.

Claims (1)

[Claims] 1. A training area is provided in the mark sheet, the cutting circuit recognizes that it is a training area, and within the training area, the type of character to be written is determined depending on the position of the entry frame, and the When performing recognition processing on the written characters, the character judgment circuit selects a font identification number that satisfies all the judgment items written in the dictionary memory, stores it in the font registration memory, and stores it in the normal area of the mark sheet after training. In recognition, the character determination circuit selects a font that satisfies determination items by searching the dictionary memory based on the font identification number registered in the font registration memory and identifies the character type. A character recognition method featuring: 2. An extraction circuit that extracts pattern information for one character from mark sheet screen information, a feature extraction circuit that extracts partial features necessary for character determination from one character pattern, and a feature counter that digitizes the degree of each of the partial features. a circuit, a dictionary memory that stores the judgment items of a sample font in order of font identification number, the character type to which the font belongs, the partial feature number to be focused on as each judgment item, its reference value, and allowable deviation value; and this dictionary memory. In a character recognition device having a character judgment circuit that collates the output of the feature counting circuit according to the judgment items and selects a font that satisfies all the judgment items, the mark sheet is filled in according to the entry frame position provided in the mark sheet. The character determination circuit includes a training area detection means for detecting a training area in which a character type is fixed, and a font registration memory, and the processing of the character determination circuit for characters written in the training area is performed based on the corresponding character written in the dictionary memory. A search is made for all fonts of the character type, and a font identification number that satisfies the determination items is selected and stored in the font registration memory, and the character determination circuit processes the normal area of the mark sheet after the training process. The character determination circuit selects a font that satisfies the determination item by searching the dictionary memory based on the font identification number registered in the font registration memory and specifies the character type. Characteristic character recognition device.