JPS62219091A - Character recognizing and deciding system - 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] [Summary] The present invention provides a character recognition device in which a recognition means is configured with a plurality of recognition units corresponding to one recognition target and each having different recognition conditions, and each recognition result is sent to a majority circuit. and enter
The result is determined by majority vote. This makes it possible to objectively determine valid results and check for misrecognition.

[Industrial application field]

The present invention relates to a character recognition determination method for preventing erroneous recognition in a character recognition device that recognizes handwritten characters, printed characters, etc.

[Prior art]

Conventional examples of character recognition devices for recognizing handwritten characters, printed characters, etc. are shown in FIGS. 3(α) and 3(b).

In the same figure (α), an image reading unit 1 optically reads a character image as multivalued image data, inputs it into a binarization circuit 2, sets a threshold value, outputs binarized data, and converts it into a binary image. Store it in the video memory 3. Regarding the image data of this binarized video memory, the position (address) 1 of the character to be recognized
The recognition target is read out to the recognition unit 4 by the definition information circuit 5 which stores definition information such as character types (alphanumeric characters and kana). This recognition unit 4 extracts the features of the recognition target, and sequentially extracts the features from a plurality of dictionaries 61.
゜621 (Verify with 53 and use the matching character as the recognized character.

Multiple Dictionary 61 # 6z + 6s is the standard for type 1 handwritten.

It is set based on the quirks of many unspecified people or the quirks of a specific person.

However, for input characters limited by definition type information,
Since the dictionary is changed sequentially, recognition time is long.

In FIG. 3B, a plurality of recognition targets are read out in parallel from the binarized video memory 3. For example, recognition target 1 is read out from the binarized video memory 3.
is read out to the recognition unit (1) 41 and compared with the dictionary (1) 7t,
Read recognition target 2 to recognition unit (2) 4 z dictionary (2) 7
2, and the recognition target 3 is read out to the recognition unit (3) 4s and compared with the dictionary (3) 73. These results are sent to the zero output circuit 8.
Synthesize and output. In this case, one recognition unit is provided with one dictionary, but a plurality of dictionaries may be provided.

In any case, the time required for the recognition section can be reduced to a fraction of that required for the recognition section.

[Problem that the invention seeks to solve]

In the conventional examples (α) and (b) above, the operator makes a judgment on the results of checking against a plurality of dictionaries in the recognition unit.

On the other hand, it is desired that the recognition unit automatically and objectively determine the results of comparison with multiple dictionaries.

SUMMARY OF THE INVENTION An object of the present invention is to provide a character recognition/determination method that automatically and objectively determines the results of comparison with a plurality of dictionaries in a handwritten character or printed character recognition unit.

[Means for solving problems]

In order to achieve the above object, according to the present invention, in a character recognition device, the recognition means is composed of a plurality of recognition units corresponding to one recognition target and each having different recognition conditions, and the recognition results of each are sent to a majority circuit. input.

FIG. 1 is a diagram explaining the principle of the present invention. That is, one recognition target is read out from the binarized video memory 3, and the recognition unit (A) 11s, CB) 112- (0)
113, recognize υ under different recognition conditions and answer A, B,
O is output, this is sent to the majority decision circuit 12, and the answer of the majority decision is output as the correct answer.

[For production]

By changing the recognition conditions of the plurality of recognition units, for example, the feature extraction and binarization conditions, similar recognition targets can be distinguished and selected by majority vote. For this reason, the recognition procedure and dictionary of the recognition unit are set in advance so that decisions can be made by majority vote within the target range limited by the definition body information circuit. In this way, the answer can be determined automatically and objectively without requiring human judgment.

〔Example〕

FIG. 2 is an explanatory diagram of the configuration of an embodiment of the present invention.

In the same figure, as in FIG. 3 (α), the image reading unit 1 optically reads a character image as multivalued image data 9, and inputs this into the binarization circuit 2 to set a threshold and convert it into binarized data. is output and stored in the binarized video memory B. From this binarized video memory 3, as explained in FIG. (0)
113. This recognition unit 1hs 11z* 113
Then, the recognition conditions are: ■ Feature extraction, such as the normal number of strokes, the shape of the outside of the direction change point, the shape of the outer contour in the left, right, up and down directions, the shape of the inner contour, etc., and ■ A combination of changes in the binarization threshold)
, respectively dictionary (A) 13t, (B) 1!12-
(0) Collation recognition with the standard form or variant stored in 153 is performed. The recognition result answers A, B, and O are sent to the majority circuit 12, where, for example, in the case of an answer of 3:0, 2:1°2:0.1:O (others are unrecognizable),
The majority answer is sent to the zero output circuit 14 as the correct answer.

In the embodiment, this method is applied by sequentially reading all characters in the input data, but if characters cannot be recognized by normal methods or characters that are likely to be misread are specified in the definition information, it is possible to apply this method. This method can also be applied and used as a misreading check method.

〔Effect of the invention〕

As explained above, according to the present invention, in the character recognition device, the recognition means is aimed at one recognition target, is composed of a plurality of recognition units each having different recognition conditions, and sends each recognition result to the majority circuit. The system inputs information and determines the result by majority vote. This makes it possible to objectively recognize the entire text or a portion of difficult-to-read characters, determine a valid result, and check for misrecognition.

[Brief explanation of drawings]

FIG. 1 is an explanatory diagram of the principle of the present invention, FIG. 2 is an explanatory diagram of the configuration of an embodiment of the present invention, and FIG. 3 (IL) is an explanatory diagram of the configuration of a conventional example. 2 is an image reading unit, 2 is a binarization circuit, 3 is a binarization video memory, 5 is a definition body information circuit, 1
11 to 113 are recognition units, 12 is a majority circuit, 151 to 1
33 is a dictionary, and 14 is an answer output circuit. Fig. 1 is an explanatory diagram of the principle of the present invention. Fig. 2 is an explanatory diagram of the configuration of an embodiment of the present invention.

Claims (1)

[Scope of Claims] Image reading means for reading handwritten characters or printed characters as binary data; a binary image memory for storing the binary data; Read the digitized data,
In a character recognition device comprising a recognition means for recognizing a character by comparing its characteristics with a dictionary, the recognition means is composed of a plurality of recognition units (11_1 to 11_3) corresponding to one recognition target and each having different recognition conditions. A character recognition determination method characterized by comprising: a majority circuit (12) for determining a result by majority vote from each recognition result from the plurality of recognition units.