JPH0362289A - Method and device for recognizing character - Google Patents

Abstract

(57) [Summary] This bulletin contains application data before electronic filing, so abstract data is not recorded.

Description

[Detailed Description of the Invention] [Object of the Invention] (Industrial Field of Application) The present invention provides an input pattern to be subjected to recognition processing.
The present invention relates to a character recognition method that performs character recognition processing using major classification processing and detailed recognition processing.

(Prior Art) Conventionally, when a character recognition device targets a character type with a large number of categories, such as kanji, it performs major classification processing for all categories to be recognized, and then The system is configured to obtain character recognition results by performing detailed recognition processing on each of the major classification candidates.

For this reason, first, major classification features for major classification processing and recognition features for detailed recognition processing are extracted from the input pattern to be recognized, and the major classification processing is performed in the major classification section. The major classification process narrows down the candidates based on the degree of similarity between the major classification features and the major classification dictionary for the major classification process, and sets major classification candidates.

The major classification section writes the major classification candidates obtained by the major classification process into the major classification candidate list memory. Next, the recognition processing unit that performs detailed recognition processing calculates the degree of similarity between the recognition features of the major classification candidates written in the major classification list memory and the recognition dictionary for recognition processing. By performing such processing for all categories, the character recognition result is determined based on the similarity obtained as a result.

(Problem to be Solved by the Invention) As described above, in the conventional method, after setting major classification candidates through major classification processing, character recognition results are obtained by performing detailed recognition processing on the set major classification candidates. . Therefore, two arithmetic units forming a major classification section that performs major classification processing and a recognition processing section that performs detailed recognition processing, and a major classification candidate list memory that stores major classification candidates are required. Therefore, the structure of the device becomes complicated, which is a big problem especially for small-sized character recognition devices.

The present invention has been made in view of the above-mentioned points, and an object of the present invention is to provide a character recognition method that allows a device to be configured with a simple circuit configuration.

[Structure of the Invention] (Means for Solving the Problems) The present invention extracts major classification features used in major classification processing and recognition features used in detailed recognition processing from an input pattern to be recognized. For each candidate category,
A dictionary means is provided to q a large classification dictionary used for the large classification process and a recognition dictionary for detailed recognition processing, and the large classification dictionary of the category to be processed and the large classification features of the extracted input pattern are used. As a result of this major classification process, it is determined whether the category should be a major classification candidate for detailed recognition processing, and based on this determination, the category is determined to be a major classification candidate. Only when it is determined that the recognition process should be performed, detailed recognition processing using the recognition dictionary of the category of the dictionary means and the recognition features of the input pattern is performed, and after this recognition processing is performed, the next step is performed. Proceed to the processing for the category to be processed, and if it is determined that the category should not be a major classification candidate, perform the major classification using the major classification dictionary of the next category to be processed. The character recognition process result is determined based on the results obtained by performing the process on all categories that are candidates for recognition by the dictionary means.

(Function) In this type of character recognition method, if as a result of performing major classification processing using a major classification dictionary for a certain category, this category can be a candidate for major classification, the recognition dictionary for this category is subsequently used. In order to perform recognition processing according to the invention, a single arithmetic unit can perform major classification processing and recognition processing.

(Example) Hereinafter, an example of the present invention will be described with reference to the drawings. FIG. 1 is a flowchart showing the processing procedure of the character recognition method according to the same embodiment. FIG. 2 is a block diagram showing an example of the configuration of a character recognition device to which the character recognition method shown in the flowchart of FIG. 1 is applied. In FIG. 2, a photoelectric conversion unit 11 optically reads an image of the entire document and converts it into an electrical signal. For photoelectric conversion 1sll,
A character cutting section 12 is connected to detect a character pattern in one character range from an image of the entire document. Connected to the character extraction section 12 is a feature extraction section 13 that extracts major classification features used in major classification processing and recognition features used in detailed recognition processing for each character pattern. The feature extraction unit 13 includes a recognition processing unit 1 that performs major classification processing and detailed recognition processing based on the characteristics of character patterns.
4 are connected. The recognition processing unit 14 is connected to a dictionary unit 15 for storing a recognition dictionary and a major classification dictionary for each category as a recognition candidate to be used when performing major classification processing and detailed recognition processing. There is. The recognition unit 14 outputs recognition results (character codes) obtained through the major classification process and the detailed recognition process.

Next, the operation of this embodiment will be explained.

First, the photoelectric conversion unit 11 optically scans a document to be processed. That is, the image of the entire document is read by illuminating the document with light and performing photoelectric conversion according to the reflected light. When the image of the entire document is read, the character cutting section 12 detects and cuts out a character pattern in one character range from the image of the document, and transfers the character pattern to the feature extraction section 13 . The feature extraction unit 13 extracts, from the transferred character pattern,
Major classification features used in major classification processing and recognition features used in detailed recognition processing are extracted. Once each feature is extracted from the character pattern, the recognition unit 14 performs character recognition processing based on this feature.

Here, the method of character recognition processing in the recognition unit 14 will be explained with reference to the flowchart shown in FIG. First, the recognition unit 14 reads out the major classification dictionary regarding the first category among the recognition candidate categories stored in the dictionary unit 15, and calculates the degree of similarity between the character pattern to be subjected to character recognition processing and the major classification feature. Calculate and find (step S1).

The recognition unit 14 compares the degree of similarity obtained here with a preset threshold, and determines that the categories corresponding to the major classification dictionary should be candidates for the major classification to be subjected to detailed recognition processing. It is determined whether or not (step S2). If the degree of similarity indicates that the chain should be a major classification candidate, the recognition unit 14 reads out the recognition dictionary for this category from the dictionary unit 15. Then, the degree of similarity between the read recognition dictionary and the recognition feature of the category is calculated and determined (step S3).

On the other hand, in step S2, if this category is not determined to be a major classification candidate, the process of calculating the similarity between the recognition dictionary of this category and the recognition features of the input pattern is not performed, and the category to be processed is Move to next. Here, if the processing has not been completed for all the categories stored in the dictionary section 15 (step S4), the recognition section 14 reads out the major classification dictionary of the category to be processed next from the dictionary section 15, and reads this dictionary. Find the degree of similarity between the major classification dictionary and the major classification features of the input pattern. Thereafter, similar processing is performed (steps S2 to S4).

In addition, in step S4, when it is determined that the processing has been completed for all categories stored in the dictionary unit 15, the recognition result is the category for which the degree of similarity indicating the maximum value obtained through detailed recognition processing has been obtained. Output as , and end the character recognition process.

Here, a specific processing example will be explained. Figure 3 shows
An example of a dictionary storage format of the dictionary section 15 is shown. That is, FIG. 3 shows a major classification dictionary and a recognition dictionary for each of the categories "sadness", "love", and "ho". Here, a major classification dictionary and a recognition dictionary are stored consecutively for each category. When performing character recognition processing, the recognition unit i4 sequentially reads out categories starting from the lowest address and executes the processing. For example, if the category "Ai" is determined to be a major classification candidate, and the categories "Ai" and "Ko" are not determined to be major classification candidates, the recognition unit 14 reads out the dictionary as indicated by the arrow in the figure. and execute the process. That is, if the recognition unit 14 reads out a major classification dictionary for a certain category and executes processing, and as a result, this category is determined to be a major classification candidate, it subsequently reads out the recognition dictionary and executes recognition processing, If it is determined that this category is not a major classification candidate, a series of processing is performed in which the major classification dictionary for the next category to be processed is read and the major classification process is executed.

Therefore, only one arithmetic unit is required for character recognition processing, and furthermore, it is possible to eliminate the need for a major classification candidate list memory for storing major classification candidates.

[Effects of the Invention] As described above, according to the present invention, large classification processing and recognition processing are not performed independently, but are performed as a series of processing, so only one calculation unit is required, and furthermore, it is possible to generate large classification candidates. Since no major classification candidate list memory is required for storage, the character recognition device can be configured with a simple circuit configuration.

[Brief explanation of drawings]

1 is a flowchart showing the processing procedure of the character recognition method according to the same embodiment; FIG. 2 is a block diagram showing an example of the configuration of a character recognition device to which the character recognition method shown in the flowchart of FIG. 1 is applied; and FIG. The figure is a diagram showing an example of a dictionary storage format of the dictionary section. 11... Photoelectric conversion section, 12... Character cutting section, 1
3...Feature extraction unit, 14...Recognition unit, 15...Dictionary unit.

Claims (2)

[Claims] (1) Detect an input pattern to be subjected to recognition processing, extract major classification features used in major classification processing and recognition features used in detailed recognition processing from this detected input pattern, and extract them for each category to be a recognition candidate. A dictionary means having a major classification dictionary used for the major classification processing and a recognition dictionary for detailed recognition processing is provided, and the said major classification dictionary of the category to be processed and the major classification features of the extracted input pattern are provided. As a result of this major classification process, it is determined whether the category should be a major classification candidate for detailed recognition processing, and based on this determination, the category is a major classification candidate. If it is determined that the input pattern should be recognized, a detailed recognition process is subsequently performed using the recognition dictionary of the category of the dictionary means and the recognition features of the input pattern, and after this recognition process is performed, the next recognition is performed. Proceed to the process for the category to be processed, and if it is determined that the category should not be a major classification candidate by the above determination, perform the major classification process using the major classification dictionary of the next category to be processed. A character recognition method that determines a character recognition processing result based on the results obtained by performing processing on all categories that are candidates for recognition by the dictionary means. (2) A detection means for detecting an input pattern to be subjected to recognition processing, and a feature for extracting major classification features used in major classification processing and recognition features used in detailed recognition processing from the input pattern detected by the detection means. extraction means; dictionary means having, for each category to be a recognition candidate, a broad classification dictionary used for broad classification processing and a recognition dictionary for detailed recognition processing; and large classification features obtained by the feature extraction means and processing. Execute a major classification process using the major classification dictionary of the dictionary means for the target category, and as a result, determine whether or not the category is a major classification candidate to be targeted for detailed recognition processing; Only when the category is determined to be a major classification candidate as a result of the determination, detailed recognition processing is subsequently performed using the recognition features and the recognition dictionary for the category, and after this recognition processing is performed, the next recognition processing is performed. Proceed to the processing for the target category, and if the category is not determined to be a major classification candidate as a result of the above determination,
a recognition means that proceeds to processing for a category to be subjected to the next recognition process; A character recognition device that determines the recognition processing result.