JPS6249593A - Character recognizing device - Google Patents

Abstract

PURPOSE:To shorten the recognition time of a handwritten character by referring to one piece of dictionary, based on a result of recognition which has referred to plural independent dictionaries in accordance with a difference of an inclination degree of the character. CONSTITUTION:When a parameter from a feature extracting part 16, which is read by a read part 10 is supplied to a dictionary retrieving part 18, independent dictionaries 20A, 20B are referred to, respectively in accordance with a difference of an inclination degree of a character, based on a designation of the first mode by a mode control part 22. Also, the smallest one of a calculated distance value is recognized as the first candidate, and in accordance with the recognition which is executed one after another, counters 24A, 24B corresponding to the dictionaries 20A, 20B, respectively are brought to an increment. Subsequently, when the second mode is set by the control part 22, only one piece of the dictionaries 20A, 20B corresponding to an inclination of a handwritten character is referred to in accordance with the counters 24A, 24B whose counting contents are large, an inclined handwritten character is recognized without referring to all the dictionaries, and the recognition time is shortened.

Description

DETAILED DESCRIPTION OF THE INVENTION [Technical Field] The present invention relates to a character recognition device for recognizing handwritten characters.

[Prior art]

In a character recognition device that recognizes handwritten characters, it is necessary to create a recognition dictionary so that it can handle the transformation of a single character.Of the 0 character transformations, the slope of 9 characters is particularly variable among individuals, and has traditionally been used to Sufficient consideration has been given to the slope of

However, in conventional character recognition devices, a dictionary for characters with an inclination and a dictionary for characters without an inclination are grouped together for each character, and the two are integrated.
Since the system was configured to search the entire dictionary to recognize input characters, there was a problem in that the recognition time was long.

〔the purpose〕

The purpose of the present invention is to be able to cope with the inclination of characters as in the past,
Another object of the present invention is to provide a character recognition device that requires shorter recognition time than the conventional one.

〔composition〕

In order to achieve this object, the character recognition device according to the present invention divides a single character recognition dictionary into a plurality of independent dictionaries depending on the degree of inclination of the character, and also uses the dictionary for character recognition as an operation mode. The dictionary has a first mode in which all are used and a second mode in which only one is used, and the dictionary used in the second mode is determined based on the recognition result in the first mode, and the The mode is configured such that the first mode is selected until information for determining a dictionary to be used in the second mode is obtained, and when that information is obtained, the mode is switched to the second mode. It is something.

[Example] An example of the present invention will be described below with reference to one drawing.

FIG. 1 is a schematic block diagram showing one embodiment of the present invention. In this figure, numeral 10 is a reading unit that optically scans an input form and reads its image. The image signal output from this reading section 10 is as follows.

The signal is subjected to binarization and noise removal processing by the binarization/noise removal section 12 and then sent to the cutting section 14 . The cutting unit 14 cuts out the image data of each character from the input binarized surface image data and sends it to the feature extracting unit 16. The feature extraction unit 16 extracts features from the input character image data.

Dictionary search unit 1 searches parameters (feature vectors) of input characters.
Send to 8.

The dictionary search unit 18 recognizes the input character by calculating the distance between the parameter of the input character and the parameter of each character registered in the recognition dictionary.

In the present invention, the recognition dictionary is divided into a plurality of independent dictionaries according to the degree of inclination of the characters. It is divided into two parts: a dictionary 20A and a dictionary 20B for slanted characters.

Although the dictionaries 20A and 20B may be physically prepared on the same storage medium, for example, they may be arranged in independent and continuous storage address spaces, and viewed from the dictionary search unit 18, they are independent dictionaries. They are clearly separated logically so that they can be treated as such.

This character recognition device has a dictionary 20A,
A first mode in which character recognition is performed using both dictionaries 20A and 20B, and a second mode in which character recognition is performed using only one of dictionaries 20A and 20B.

This operation mode switching control is performed by the mode control section 22. In connection with this mode control, dictionary 2OA,
2. Counters 24A and 24B corresponding to each OB
is provided. These counters 24A and 24B are
It is incremented by the dictionary search unit 18 during operation in the first mode, and frequency information for selecting a dictionary to be used in the second mode is obtained. Based on this information, the mode control unit 22 performs operation mode switching control and dictionary selection.

Next, the operation of this character recognition device will be explained. Initially, the mode control unit 22 sets the operation mode to the first mode.

The input form is read by the reading unit 10 , one-character image data is sequentially extracted by the extraction unit 14 and input to the feature extraction unit 16 , and its parameters are input to the dictionary search unit 18 .

Since the first mode is specified by the mode control unit 22, the dictionary search unit 18 uses the dictionary 20A.

The distance between the parameter of the character registered in both 20B and the parameter of the input character is calculated, and! & Search for the first candidate character at a short distance and output its character code as a recognition result. Further, the dictionary search unit 18 increments the counter 24A by 1 when the first candidate character is found in the dictionary 20A, and increments the counter 24B by 1 when the first candidate character is found in the dictionary 20B. however,
At the start of operation, counters 24A and 24B are cleared.

In this way, as character recognition progresses, each dictionary 2OA,
The number of times the first candidate is found from 20B, that is, frequency information is accumulated in counters 24A and 24B.

The mode control unit 22 monitors the difference between the values of the counters 24A and 24B, and when the difference exceeds a predetermined value, it determines that information necessary to switch the operation mode to the second mode has been collected. Then, the second
In addition to specifying the mode of the counters 24A and 24B,
The dictionary 2OA or 20B corresponding to the larger value of is designated as the dictionary to be used.

From this point on, the dictionary search unit 18 operates in the second mode and recognizes input characters using only one of the words 1F 20A or 20B designated by the mode control unit 18.

Although the slant of handwritten letters varies greatly from person to person, the slant of the letters of a particular writer tends to be almost constant.

Therefore, even if you switch to a mode that selects and uses one dictionary, the correct reading rate will hardly decrease, and
Since the capacity of the dictionary is halved compared to when both dictionaries are used, the recognition time is also halved.

Note that in this embodiment, when the difference between the values of the counters 24A and 24B exceeds a predetermined value, the operation mode is switched to the second mode. The mode may be switched at the time when the recognition is completed.

Alternatively, the degree of inclination of one character may be classified into three or more levels, and an independent dictionary may be provided for each classification.

Next, other embodiments of the present invention will be described.

The block configuration of the device may be the same as that shown in FIG. 1, so in the second embodiment of the present invention, which will be described with reference to FIG.

A first candidate character having a minimum distance from the input character during the first mode period and a second candidate character having a distance from the input character are determined. If the first candidate character and the second candidate character are selected from different dictionaries, the corresponding counter 24A or 24B is incremented by 1 as in the first embodiment.

However, when the first candidate character and the second candidate character are selected from the same dictionary, the increment amount of the counter 24A or 24B corresponding to the dictionary from which the first candidate character is selected is increased by 1, for example, by 2. .

In this way, by increasing the amount of increment when both candidate characters are selected from the same dictionary, the number of characters required to obtain the information necessary for determining the dictionary to be used in the second mode can be increased from the number of characters in the first embodiment. can be reduced more than in the case of

Other than this, the second embodiment is the same as the first embodiment.

According to the third embodiment of the present invention, in the first mode, when the first candidate character and the second candidate character are selected from different dictionaries, the dictionary search unit 18 selects the input character of each candidate character. Find the difference in distance between the
As in the case where both candidate characters are selected from the same dictionary, the counter 24A.

Increase the 24B increment amount. The configuration other than this is the same as that of the second embodiment.

According to this embodiment, the dictionary to be used in the second mode can be determined with an even smaller number of characters than the second embodiment.

According to the fourth embodiment of the present invention, the output data (character code or redirect code) of the dictionary search section 18 is input to the mode control section 22 . and.

If a rejection occurs during operation in the second mode and a reject code is output from the dictionary search unit 18, the mode control unit 22 switches the operation mode to the first mode again and clears the counters 24A and 24B. The operations thereafter are the same as those after the start of the operations.

In this way, the success rate can be further improved.

Although the embodiments of the present invention have been described above, the present invention is not limited to each embodiment, and can be implemented with appropriate modifications.

〔effect〕

As explained above, according to the present invention, it is possible to achieve the effect of realizing an excellent character recognition device in which the recognition time for handwritten characters is significantly shortened compared to the conventional one.

[Brief explanation of drawings]

FIG. 1 is a schematic block diagram of a character recognition device according to the present invention. 10... Reading section, 12... Binarization/noise removal section. 14... Extraction section, 16... Feature extraction section. 18... Dictionary search unit, 20A, 20B... Dictionary. 22...Mode control section, 24A, 24B...Counter. 1st Figure 4B

Claims (3)

[Claims] (1) In a character recognition device that recognizes handwritten characters,
The character recognition dictionary is divided into multiple independent dictionaries according to the degree of inclination of the characters, and there are two operating modes: a first mode in which all of the dictionaries are used for character recognition, and a second mode in which only one is used. 2 modes, the dictionary used in the second mode is determined based on the recognition result in the first mode, and the operation mode determines the dictionary used in the second mode. A character recognition device characterized in that the first mode is selected until information for the character is obtained, and when the information is obtained, the character recognition device is switched to the second mode. (2) The character recognition device according to claim 1, wherein when a rejection occurs in the second mode, the operation mode is switched from the second mode to the first mode. (3) The information is information related to the frequency with which candidate characters are selected for each of the dictionaries, and if a difference in frequency of more than a certain value is obtained, the operation mode changes from the first mode to the second mode.
3. The character recognition device according to claim 2, wherein the character recognition device can be switched to an operation mode.