JPH07104911B2 - Pattern recognition device - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION [Industrial field of application] The present invention provides a belonging category recognition factor that includes a numerical value that increases or decreases at least according to the magnitude of the appearance frequency of each category in the recognition process for determining the belonging category of an input pattern. The present invention relates to a pattern recognition device that recognizes a category to which an input pattern belongs by using a corresponding number of standard templates and starts pattern recognition.

Some character recognition devices require a recognition of which character category the input character pattern belongs to.

Such recognition is intended to contribute to the improvement of the character recognition rate, but the conventional recognition technique is not always satisfactory, and further improvement efforts are required.

This situation also applies to other pattern recognition.

[Conventional technology]

A plurality of standard templates are used for each category to recognize a character category in a conventional pattern recognition device, for example, a handwritten character recognition device. This is to prevent deterioration of character recognition accuracy due to deformation of input characters for each category to be recognized and to improve the recognition rate.

As a technique of pluralizing such standard templates,
There are a method in which the same number of standard templates are prepared for each category, a method in which recognition is performed using a learning sample prepared in advance, and the number of standard templates corresponding to categories having a poor recognition rate is increased.

[Problems to be solved by the invention]

In the standard template duplication technique as described above, the standard template that is compared with the input character when recognizing the category of the input character is pluralized regardless of the frequency of use of the character. Character recognition accuracy does not necessarily increase.

[Means for solving problems]

The present invention provides a pattern recognition device that solves the above-mentioned problems, and the means is a pattern recognition device that uses a standard template in the recognition process for determining the category to which an input pattern belongs, and at least the appearance frequency of each category. The standard template output means for outputting the number of standard templates corresponding to the category recognition factor including the numerical value that increases or decreases according to the size of And an identification means for outputting.

[Action]

According to the apparatus of the present invention, the recognizing process for determining the belonging category of the input pattern includes at least the belonging category determining factor including a numerical value that increases or decreases depending on the magnitude of the appearance frequency.

Thereby, the recognition accuracy of the input pattern can be improved.

〔Example〕

 Hereinafter, embodiments of the present invention will be described with reference to the accompanying drawings.

FIG. 1 shows a block diagram of an embodiment of the present invention. In this figure, 1 is an observation unit that reads a sheet on which a pattern to be recognized such as characters and figures is read and photoelectrically converts it, 2 is a feature extraction unit that extracts the features from the pattern input from the observation unit, and 3 is a standard template Is a recognition dictionary storing 4), and 4 is an identification unit that compares the features of the input pattern from the feature extraction unit 2 with the standard template from the recognition dictionary 3 and outputs a candidate category.

A standard template is determined and stored in the recognition dictionary 3 as described below. It is assumed that the recognition target categories input via the observation unit 1 are _N categories of C ₁ , C ₂ , ..., C _N , for example, 26 categories if they are alphabets, and the appearance frequency of each category is P ₁ , P ₂ , ・ ・ ・ P _N Then, the standard template number mi of the category Ci is determined as mi = f (Pi), and those standard templates are selected (see FIG. 2).
Is stored in the recognition dictionary 3 in advance. One example used as the function f determined in this way is that mi is mi = [(M−N) × Pi] +1 where M is the maximum number of all standard templates, where [a] is a Gaussian symbol. Yes, it is the maximum integer that does not exceed a.

Is an increasing function that is determined as.

The recognition of categories in the device configured in this way is as follows.

The features of the recognition target pattern output from the observing unit 1 are extracted by the feature extracting unit 2.

The extracted features are sent to the identification unit 4, where the categories to which the standard template group determined to be similar to the features by being compared with the standard templates read from the recognition dictionary 3 belong are output as candidate categories.

In this way, the number of standard templates that increases / decreases in accordance with the magnitude of the appearance frequency is used in recognizing the category to which each input pattern belongs. Therefore, it greatly contributes to the recognition of an input pattern having a high appearance frequency with high accuracy. In addition, the storage capacity is reduced as compared with the case where the number of standard templates is uniformly set for each category.

It should be noted that in the above embodiment, an example in which the appearance frequency is used as the belonging category recognition factor has been described, but the factor may be determined by adding the recognition rate and the like to the appearance frequency. That is, when a plurality of standard templates having various modifications for each category are stored as a set like the device of the present invention, the recognition rate increases as the number of standard templates increases, so the frequency of appearance The number of standard templates may be increased or decreased according to the size of each, and a larger number of standard templates may be allocated to a category with a low recognition rate.

Although the number of standard templates is statically determined, the number of standard templates may be dynamically determined by the factor in the process of recognizing the input pattern. This will contribute to the improvement of recognition accuracy.

〔The invention's effect〕

As described above, according to the present invention, it is possible to recognize a category that immediately responds to an input pattern, which greatly contributes to improvement in the recognition accuracy of the input pattern.

[Brief description of drawings]

FIG. 1 is a block diagram of an example of the main configuration of the present invention, and FIG. 2 is a diagram showing an example of a standard template group for each category stored in the FIG. 1 recognition dictionary. In the figure, 1 is an observation unit, 2 is a feature extraction unit, 3 is a recognition dictionary, and 4 is an identification unit.

Claims (2)

[Claims] 1. A pattern recognition apparatus using a standard template in a recognition process for determining a category belonging to an input pattern, the pattern recognition apparatus comprising a number according to a category recognition factor including a numerical value that increases or decreases at least according to the magnitude of appearance frequency of each category. A pattern recognition apparatus comprising: a standard template output means for outputting a standard template; and an identification means for comparing an input pattern with a standard template group output corresponding thereto and outputting a candidate category. 2. The pattern recognition according to claim 1, wherein the standard template output means is a dictionary that stores a number of standard templates that increase or decrease in accordance with the magnitude of the appearance frequency of each category. apparatus.