JPS6121583A - Pattern recognizer - Google Patents

Abstract

PURPOSE:To reduce significantly the time necessary for sorting by deciding a category group to which an input pattern belongs according to a similarity degree of features of the input pattern and averaged those features. CONSTITUTION:The number of categories by category group of the 2nd feature dictionary 7 and that of the 1st feature dictionary 1 are set to ''10'' and ''2,000'' types, respectively, while types of standard features by category, that is, types of features extracted in a feature extraction part 2 is set to ''10'' types. Moreover, the number of features used in the 1st recognition part 3' and that in the 2nd recognition part 6 are set to ''1'' and ''10'' types, respectively. At such a time, the recognition part 3' calculates a similarity degree 2,000 times, selects high-order two categories from the top among these degree, then selects about high-degree 20 categories similar to the input pattern from the top. Afterwards the category of the input pattern can be decided by the similarity degree calculation about 20X10 times.

Description

DETAILED DESCRIPTION OF THE INVENTION [Field of Industrial Application] The present invention relates to a pattern recognition device that improves recognition speed.

Pattern Recognition Device For example, in a character recognition device, a feature dictionary is usually provided with features extracted from standard character patterns for each category (type of character), and input character butter l.
That is, the category that has the highest degree of similarity to the feature extracted from the character pattern to be recognized is selected from the feature dictionary, and this is determined as the category of the character pattern to be recognized.

Further, as the above-mentioned features, for example, in Chinese characters, a plurality of types of features are used, such as the shape of projection, the number of closed loops, and the number of separation strokes.

By the way, in order to select a category that has a high degree of similarity, if we calculate the degree of similarity between the features extracted from the character pattern to be recognized and the features for each category stored in the feature dictionary, for example: Type of feature is 1
If there are 0 types and 2000 categories, it is necessary to calculate the similarity 20,000 times.

Therefore, to reduce the number of similarity calculations.

Recognition is divided into two stages: primary recognition and secondary recognition.

- In the next recognition, multiple categories with high similarity are selected as candidates through rough recognition using only some features,
In secondary recognition, a pattern recognition device has been proposed that selects the category with the highest degree of similarity to the input pattern from among the multiple categories selected in -order recognition through precise recognition using all features. ing.

At this time, by efficiently selecting multiple categories in -order recognition, 1. It is desired to further improve the recognition speed.

[Conventional technology]

Figure 2 shows a pattern recognition unit that aims to reduce the number of similarity calculations between the features extracted from the input pattern and the features stored in the feature dictionary! This is a conventional example.

In the figure, 1 is a feature dictionary that stores predetermined multiple types of features extracted from standard character patterns for each category as a feature dictionary, and 2 is a feature extraction that extracts the same predetermined multiple types of features from input patterns. Part 3 is a predetermined part of the features of the input pattern extracted by the feature extraction part 2.

The first recognition unit (
-Jiken&i! 4 is a selection unit that selects categories up to a predetermined top rank in the order of similarity obtained by the first recognition unit 3; 5 is a selection unit that selects the standard features of each category selected by the selection unit 4 in the feature dictionary 1; a buffer that is retrieved from and stored in
6 is the feature of the input pattern extracted by the feature extraction unit 2;

This is a second recognition unit (secondary recognition m) that calculates the degree of similarity between each of a predetermined number of categories stored in the buffer 5 and standard features, and determines the category with the highest degree of similarity as the category of the input pattern.

In the above configuration, for example, the number of categories stored in the feature dictionary 1 is 2000 types, and the types of standard features for each category stored in the feature dictionary 1, that is, the types of features extracted in the feature extraction unit 2 are 10 types. , the number of features used in the first recognition unit 3 is one type, and the number of features used in the second recognition unit 6 is
Assuming that the number of features used is 10 and the top 20 categories are selected in the selection unit 4, the first recognition unit 3 calculates the degree of similarity 2000 times, and the second recognition unit 6 calculates the degree of similarity 2000 times. A total of 2200 similarity calculations will be performed, including 9 calculations (20X 10) times.
It can be seen that the number of calculations can be significantly reduced compared to 0.000 times.

[Problem that the invention seeks to solve]

However, in the configuration described above, the 9 selection section 4 needs to select the top 20 from among the similarities determined by the first recognition section for the 2000 categories through a sorting process or the like. , there is a problem that the processing for this requires a long time.

[Means for solving problems]

The pattern recognition device according to the present invention includes a first feature dictionary that stores standard features for each category, and a second feature dictionary that stores categories similar to each category and average features for each category group. a dictionary, a feature extraction unit that extracts features of the input pattern, and a category group to which the input pattern belongs is determined based on the degree of similarity between the extracted features of the input pattern and an average feature stored in the second feature dictionary. and standard features of categories included in the category group determined by the first recognition unit among the features of the extracted input pattern and the standard features stored in the first recognition dictionary. The above problem is solved by including a second recognition unit that determines the category of the input pattern based on the degree of similarity between the input pattern and the input pattern.

[Effect]

That is, in the present invention, in addition to a first feature dictionary that stores standard features for each category, a second feature dictionary that stores 8491 categories for each category and the average feature for each category group is provided. For example, by determining the category group to which the input pattern belongs based on the similarity between the features of the input pattern and the average feature stored in the second feature dictionary, If the number of categories is 10, the top 20 categories that are similar to the input pattern are selected by selecting the top 2 categories from the similarity determined by the first recognition unit. can be selected, and the time required for sorting and sorting can be significantly reduced compared to the conventional example.

〔Example〕

Next, the gist of the present invention will be specifically explained using an embodiment shown in FIG.

FIG. 1 is a block diagram showing the configuration of an embodiment of the present invention, in which the same reference numerals as in FIG. 2 refer to the same objects, and 7
is a second feature dictionary that stores categories similar to each category and average features for each category group.

3' is a first recognition unit that determines the category group to which the input pattern belongs based on the degree of similarity between the features of the input pattern extracted by the feature extraction unit 2 and the average features stored in the second feature dictionary; Further, 4' is a selection section that selects a category group up to a predetermined value in the order of similarity obtained by the first recognition section 3'; This is a buffer that extracts standard features from the first feature dictionary 1 and stores them.

In the above configuration, in addition to setting the number of categories for each category group stored in the second feature dictionary'l to 10,
Similar to the embodiment described above, the number of categories stored in the first feature dictionary 1 is 2000, and the types of standard features for each category are stored in the first feature dictionary 1.

In other words, the type of feature extracted in the feature extraction unit 2 is set to 1.
0 types, the number of features used in the first recognition unit 3' is 1
Assuming that the number of types and features used in the second recognition unit 6 is 10, the first recognition unit 3° calculates the degree of similarity 2000 times as in the conventional example.

By selecting the top two category groups from among them, it is possible to select the top 20 categories that are similar to the input pattern, and then the second recognition unit 6 calculates the similarity about (20 x 10) times. The category of the input pattern can be determined by calculation.

That is, the number of similarity calculations is approximately the same as in the conventional example, but compared to the conventional example, the time required for sorting can be significantly reduced by selecting one selection section 4'.

[Effects of the Invention] As explained above, according to the present invention, the recognition speed of a pattern recognition device can be greatly improved.

[Brief explanation of drawings]

FIG. 1 is a configuration diagram of an embodiment of the present invention. FIG. 2 shows a configuration diagram of a conventional example. In the figure. 1 is the first feature dictionary, 2 is the feature extraction unit, 3° is the first recognition unit, and 6 is the second recognition unit.・7 is the second feature dictionary. Figure 1

Claims (1)

[Claims] A first feature dictionary that stores standard features for each category, a second feature dictionary that stores categories similar to each category and average features for each category group, and extracts features of input patterns. a first recognition unit that determines a category group to which the input pattern belongs based on the degree of similarity between the extracted features of the input pattern and the average feature stored in the second feature dictionary; The input pattern is determined based on the degree of similarity between the extracted input pattern feature and the standard feature of the category included in the category group determined by the first recognition unit among the standard features stored in the first recognition dictionary. A pattern recognition device characterized by comprising: a second recognition section that determines a category.