JPH06111078A - Character recognition system - Google Patents

Abstract

PURPOSE:To always keep a parameter to adjust the collation range of a word dictionary at a proper value, to improve a recognition rate, and to accelerate processing speed in a character recognition system. CONSTITUTION:The character recognition system which performs the collation of plural character candidates extracted from image data with a stored word, and estimates a correct answer character, and recognizes a character from the image data is provided with a parameter control part 6 which dynamically controls the setting of the parameter to adjust the collation range of the word dictionary used in a collation operation for the character candidate based on the correct answer rate of an estimated correct answer character and estimation processing time, thereby, a parameter value to adjust the collation range of the word dictionary can be always kept at the proper value.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a character recognition processing technique for recognizing a target character from image data such as an OCR (optical character reader), and more particularly to a correct character recognition rate and The present invention relates to a character recognition post-processing control system suitable for improving the recognition processing speed.

[0002]

2. Description of the Related Art At present, a character recognition system, such as an OCR (optical character reader), has been developed for recognizing a target character from image data using a computer. The character recognition system recognizes a character by extracting a reference pattern similar to the input pattern by collating an input pattern of a character to be recognized with a reference pattern registered in advance. Regarding such a character recognition system, for example, "Electronic Information and Communication Handbook" (19) edited by the Institute of Electronics, Information and Communication Engineers
1988, published by Ohmsha), pages 1114 to 1124.
Page.

Generally, the collating operation is performed twice.
Some reference patterns that are similar to the input patterns are extracted as character candidates when extracting character candidates. Furthermore, when estimating correct characters from these character candidates, matching with the words stored in the word dictionary is performed. Is performed. The matching between the character candidates and the words stored in the word dictionary is performed by the character recognition post-processing device.

Conventionally, in this character recognition post-processing device, the proper value of the parameter for adjusting the collation range of word collation has been fixedly set to a value obtained by evaluation measurement in an environment assumed in advance. Therefore, with respect to changes due to the quality of input image data and differences in character writers, word matching is performed with this parameter value not set to an appropriate value. Conventionally, in order to maintain such parameters at appropriate values, resetting by the operator is necessary. Similarly, even after resetting, it is necessary to reset again in response to changes in the input image data, which puts a burden on the operator.

[0005]

The problem to be solved is that in the conventional technique, the parameter for adjusting the collation range of the word dictionary cannot be maintained at an appropriate value without intervention of an operator. An object of the present invention is to solve these problems of the prior art, automate the setting of parameters used for adjusting the matching range of the word dictionary, reduce the load on the operator, and correct the recognition rate of characters and the recognition processing speed. It is to provide a character recognition system capable of improving.

[0006]

In order to achieve the above object, the character recognition system of the present invention comprises: (1) a character candidate extraction unit for recognizing a character from image data and extracting a plurality of similar character candidates. The character candidate extraction unit is provided with a word collation unit that collates each character candidate extracted by the character candidate extraction unit with a word stored in a word dictionary and estimates a correct character.
In a character recognition system that performs character recognition from image data, the value of a parameter that adjusts the matching range of the word dictionary used for matching the character candidates by the word matching unit with the words stored in the word dictionary is set to the word matching. The present invention is characterized in that a parameter control unit for dynamically setting the parameter value for dynamically adjusting the collation range of the word dictionary based on the estimation result of the correct answer character of the copy section and keeping the parameter value at an appropriate value is provided. Also,
(2) In the character recognition system according to (1) above,
The parameter control unit includes a first set value calculation unit that calculates a value of a parameter that is dynamically set based on the correct answer rate of the correct answer character estimated by the word matching unit. (3) In the character recognition system described in (1) above, the parameter control unit calculates the value of the parameter that is dynamically set based on the correct processing character estimation processing time of the word matching unit. It is characterized by comprising a value calculation unit. (4) In the character recognition system according to (1) above, the parameter control unit calculates the value of the parameter that is dynamically set based on the correct answer rate of the correct characters in the word matching unit and the estimated processing time. It is characterized by comprising three set value calculation units.

[0007]

In the present invention, each time the appropriate value of the parameter for adjusting the collation range of the word dictionary changes, the ratio of correctly recognized characters and the processing time required for word collation are calculated for each parameter. Then, they are automatically corrected to appropriate parameter values when the system user has desired values. In this way, the parameters for adjusting the matching range of the word dictionary can be dynamically re-set, and the parameters can always be kept at appropriate values, improving the ratio of correctly recognized characters and the processing time required for word matching. It will be possible. Further, it is not necessary for the operator to perform resetting, and the burden on the operator can be reduced.

[0008]

Embodiments of the present invention will now be described in detail with reference to the drawings. FIG. 1 is a block diagram showing a first embodiment of a configuration relating to the present invention of a character recognition system of the present invention.
In the figure, 1 is a document reading device for reading image data, 2 is a pre-processing device for cutting out desired characters from the image data read by the document reading device 1, 3
Is a character recognition device that recognizes the characters cut out by the pre-processing device 2 and outputs similar character candidates, 4 is a character recognition post-processing device that extracts correct characters from the character candidates output by the character recognition device 3, 7 Is an operator operating device that includes a display, a keyboard, and the like, and allows an operator to correct erroneously recognized characters. The character recognition post-processing device 4 matches the character candidates output by the character recognition device 3 with the words stored in the word dictionary, and estimates the correct character.
According to the present invention, the parameter control unit 6 dynamically controls the setting of the parameter for adjusting the matching range of the word dictionary used for the matching operation of the word matching 5.

The operation of the character recognition system of this embodiment will be described below. The image data read by the form reading device 1 is cut out by the preprocessing device 2 for each character, and the character recognition device 3 recognizes each character and outputs a recognition candidate character string. Then, the character recognition post-processing device 4 uses the word matching unit 5
At this time, the recognition candidate character string output from the character recognition device 3 is collated with the word dictionary, and the result is compared with the operator operation device 7
Pass to. In the operator operating device 7, the collation result is displayed on the display, and the operator compares the collation result with the form to correct the erroneously recognized character. This correction information is passed to the parameter control unit 6.

In the parameter control unit 6, the word collation unit 5 is based on the erroneously recognized character correction information from the operator operating device 7.
Control the parameters used in. In the parameter control unit 6, the parameters set by the word matching unit 5 are passed to the word matching unit 5 and used in the next form. The control operation of the parameter control unit 6 will be described below with reference to FIG.

FIG. 2 is a flow chart showing an embodiment of the processing operation according to the present invention of the parameter control unit in FIG. The parameter control unit 6 calculates the ratio of correctly recognized characters based on the erroneously recognized character correction information from the operator operating device 7 (step 61). This calculation is performed for each read form. The ratio of characters that were correctly recognized when the parameter value was set to the optimal value last time was used as the base axis, and the ratio of a certain value from this base axis was used as the threshold value. It is determined whether the value is exceeded (step 62).

If the determination result exceeds the threshold value, the word collation unit 5 obtains the collation result for each parameter, and the ratio of the characters correctly recognized in each parameter is calculated from this collation result (step 63). The parameter value having the highest ratio of the result is selected as an appropriate value, and the appropriate value is passed to the word matching unit 5 (step 64). Then, the next threshold value is set with the ratio of correctly recognized characters when the parameter is an appropriate value as the base axis (step 65). In step 62, if the determination result does not exceed the threshold value, the appropriate value is not changed and the parameter value is not set.

FIG. 3 is an explanatory diagram showing an example of the concept of the threshold value used in the processing operation of the parameter control unit in FIG. In the figure, 31 is the ratio (0% to 100%) of characters that can be correctly recognized, which is calculated based on the misrecognized character correction information from the operator operating device 7 in FIG.
2 is a base axis indicating the ratio of characters that were correctly recognized when the parameter value was set to the optimum value last time, and 33 and 34 are ratios obtained by adding or subtracting a certain value from the base shaft 32, which are used as evaluation scales for appropriate values. It is a threshold. The parameter control unit 6 of FIG. 1 selects a new appropriate value and the basic axis when the ratio of correctly recognized characters exceeds the shaded area in the drawing, that is, the threshold values 33 and 34. 32 is set.

In the character recognition system shown in FIG. 1, the parameter control unit 6 sets the value of the parameter dynamically set to
Using the first set value calculation unit of the present invention not shown,
The value is calculated based on the correct answer rate of the correct answer character of the word matching unit 5, but the value of the parameter is determined by using the second setting value calculating unit of the present invention as shown in FIG. It can also be calculated based on the estimated processing time.

FIG. 4 is a block diagram showing a second embodiment of the structure relating to the present invention of the character recognition system of the present invention. In the figure, 1 is a form reading device that reads image data, 2 is a preprocessing device that cuts out a desired character from the image data read by the form reading device 1, and 3 is a character that is cut out by the preprocessing device 2 Then, the character recognition device 4a for outputting similar character candidates is a character recognition post-processing device for extracting correct characters from the character candidates output by the character recognition device 3, and 7 is provided with a display, a keyboard, etc. This is an operator operating device that corrects erroneously recognized characters.

The character recognition post-processing device 4a collates the character candidates output by the character recognition device 3 with the words stored in the word dictionary, and in the present invention, a word collating unit 5 for estimating the correct character. Relatedly, it is configured by a parameter control unit 6a that dynamically controls the setting of the parameter that adjusts the matching range of the word dictionary used for the matching operation of the word matching 5. The parameter control unit 6a also includes a second setting value calculation unit 8 that calculates the value of the parameter that is dynamically set based on the estimated processing time of the word matching unit 5.

The character recognition operation in the character recognition system of the second embodiment is the same as the operation of the first embodiment described with reference to FIG. 1, but the character recognition system of the second embodiment is the same. Then, the value of the parameter for adjusting the collation range of the word dictionary set by the parameter control unit 6a is calculated by the second setting value calculation unit 8. That is, in the parameter control unit 6a, the second set value calculation unit 8 calculates the time required for the estimation process of the word matching unit 5 for each of the read forms. Then, the estimated processing time when the parameter value was set to the optimal value last time was set as the base axis, and a certain value was added or subtracted from this base axis to set the threshold value, and this threshold value was compared with this calculation result. Based on this, an appropriate value for the parameter is selected.

As described above with reference to FIGS. 1 to 4, in the character recognition systems of the first and second embodiments, the parameter for adjusting the collation range of the word dictionary is dynamically set. It can always be set to an appropriate value. This makes it possible to deal with changes in input form conditions, such as image data quality and differences in writers, preventing a decrease in the ratio of correctly recognized characters and an increase in the processing time required for post-processing. be able to. Further, it is possible to reduce the burden on the operator for setting the parameters.

The present invention is not limited to the embodiment described with reference to FIGS. 1 to 4, and various modifications can be made without departing from the scope of the invention. For example, in the first and second examples, as the evaluation scale of the appropriate value, the ratio of the correctly recognized characters obtained by the first setting value calculation unit and the second setting value calculation unit were obtained. Although the estimated processing time of the word matching unit was used individually, a combination of this processing speed, the ratio of correctly recognized characters, and the processing speed was obtained by the third set value calculation unit, and used as an evaluation scale of an appropriate value. You may use.

[0020]

According to the present invention, in the character recognition system, the parameter for adjusting the collation range of the word dictionary can be kept at an appropriate value at all times without the intervention of the operator, and the operator's load is reduced and the correct value is obtained. It is possible to improve the character recognition rate and the recognition processing speed.

[0021]

[Brief description of drawings]

FIG. 1 is a block diagram showing a first embodiment of a configuration according to the present invention of a character recognition system of the present invention.

FIG. 2 is a flowchart showing an embodiment of the processing operation of the parameter control unit in FIG. 1 according to the present invention.

FIG. 3 is an explanatory diagram showing an example of the concept of threshold values used in the processing operation of the parameter control unit in FIG.

FIG. 4 is a block diagram showing a second embodiment of the configuration related to the present invention of the character recognition system of the present invention.

[Explanation of symbols]

 1 form reading device 2 pre-processing device 3 character recognition device 4 character recognition post-processing device 5 word collation unit 6, 6a parameter control unit 7 operator operation device 8 second set value calculation unit 31 ratio of correctly recognized characters 32 base axis 33, 34 threshold

 ─────────────────────────────────────────────────── ─── Continuation of the front page (72) Inventor Hirotaka Fuchizawa 1-26-5 Toranomon Minato-ku, Tokyo NTT Data Communications Co., Ltd.

Claims (4)

[Claims] 1. A character candidate extracting means for recognizing a character from image data and extracting a plurality of similar character candidates, and each character candidate extracted by the character candidate extracting means and a word stored in a word dictionary. In a character recognition system that includes a word collating unit that collates with the word and estimates a correct character, and performs character recognition from image data, the word candidate stored by the word collating unit and the word stored in the word dictionary. The value of the parameter for adjusting the matching range of the word dictionary used for matching with, based on the estimation result of the correct character of the word matching means, dynamically set, the parameter for adjusting the matching range of the word dictionary A character recognition system characterized by comprising a parameter control means for keeping the value at an appropriate value at all times. 2. The character recognition system according to claim 1, wherein the parameter control means calculates the value of the dynamically set parameter based on the correct answer rate of the correct character estimated by the word matching means. A character recognition system comprising a first setting value calculation means. 3. The character recognition system according to claim 1, wherein the parameter control means calculates a value of the parameter dynamically set based on an estimated processing time of the correct character of the word matching means. A character recognition system comprising a second set value calculating means. 4. The character recognition system according to claim 1, wherein the parameter control means sets the value of the parameter dynamically set based on the correct answer rate of the correct answer character and the estimation processing time of the word matching means. A character recognition system comprising a third setting value calculating means for calculating.