JPH02230488A - Character recognizing device - Google Patents

Abstract

PURPOSE:To dispense with registering work by a person in advance by automatically generating a limited character kind table targeted to be recognized by providing a character kind fetching part to fetch all character kinds out of a word group and to store them in a character kind storage part. CONSTITUTION:When a character recognition part 2 outputs a candidate character 3 to a word collation part 4, no character kind other than a targeted one is fetched as a candidate as referring to the character kind storage part 5, and the word collation part 4 collates the candidate character 3 with the word group stored in a word storage part 6, and outputs the character as a candidate word 7. Here, the character kind fetching part 8 is comprised between the word storage part 6 and the character kind storage part 5, and picks up the character kind comprising each word and forms the limited character kind table corresponding to the word group into the character kind storage part 5. Consequently, it is possible to easily judge an appropriate candidate word at the word collation part 4. In such a way, the generation of a table for the limitation of the character kind can be automatically performed, and no registering work by the person is required.

Description

[Detailed Description of the Invention] [Industrial Field of Application] The present invention relates to a character recognition device, and particularly relates to a character recognition device in which groups of words to be recognized according to various purposes such as prefecture names are stored in advance. The present invention relates to a character recognition device having a word matching function for checking whether a word obtained by combining character recognition results exists in a word group.

[Conventional technology]

Conventionally, this type of character recognition device has a structure in which the character recognition section and the word matching section are logically separate, and the character recognition section selects valid candidates as recognition results from among all the character types to be read. The operation was to select it and send it to the word matching section.

In addition, the recognition unit has its own table for limiting character types.
Some of these tables were created by narrowing down the types of characters to be recognized and selecting candidates, or by manual operations when creating these tables.

[Invention or problem to be solved]

The above-mentioned conventional character recognition devices have the drawbacks that, when recognizing all character types, the recognition unit takes a long time to recognize the large number of character types, and that inappropriate candidate characters are sent to the word matching unit. The disadvantage was that it took extra time to hammer out the words.

Furthermore, in the case of a device in which the recognition unit has a table for limiting character types, the time required for recognition and word matching can be shortened, but many manual operations are required to create the table. For example, when registering n words as words for matching, a person would have to pick up the character types used in all n words and register these character types in a table. . For this reason, as the number of words n increases, there is a drawback that it becomes more time-consuming for humans to check the type of characters.
This also resulted in check errors and a drop in recognition rate.

[Means to solve the problem]

The present invention includes a pure word storage unit that stores a group of target words in advance, a character type storage unit that stores a character type that is a recognition target in advance,
a character recognition unit that recognizes and identifies input character images and outputs candidate characters limited to character types stored in the character type storage unit; and a character recognition unit that outputs candidate characters limited to character types stored in the character type storage unit; In a character recognition device comprising a word matching unit having a function of checking whether a word combining candidate characters exists in a group of words stored in the word storage unit, the word matching unit is registered in the word storage unit. The present invention includes a character type extraction section which picks up all the character types used in the Chinese language group and stores them in the character type storage section.

〔Example〕

Next, the present invention will be explained with reference to the drawings.

FIG. 1 is a block diagram of a reading device of the present invention. The character image] is an image of a character manually generated from a medium such as paper, and the character recognition unit 2 recognizes what character is in the character image and uses the result as a candidate character 3 for reproduction. Output to the matching section 4. At this time, character 32 recognition part 2
has a function of not selecting character types other than the target character types as candidates without referring to the character type storage unit 5 that stores character types to be recognized. The word collation section 4 temporarily stores a plurality of candidate characters 3 outputted from the character edict and identification section 2, and stores words that can be formed by combining the candidate characters 3 into 7 in the word storage section 6. It has a function of comparing word groups and outputting matching or similar words as candidate words 7.

On the other hand, a character type extraction unit 8, which is the core of the present invention, is configured between the 111 word storage unit 6 and the character type storage unit 5. It has a function of picking up the character types constituting the word group and forming a limited character type table corresponding to the word group in the character type storage section 5.

In order to clearly demonstrate the effects of the present invention, the operations of each part will be explained below, taking the case of recognizing prefecture names as an example.

Prior to recognition, all prefecture names (such as ``Tokyo'', ``Kanagawa Prefecture'', etc.) are registered as a word group in the word storage section 6. Next, the second section 8, which picks up character types, is activated. Then, select the character types used for all prefecture names and create a limited table of prefecture names in the character type storage section 5g) 111G.

When the recognition operation starts, character images 1 are input one after another, or as an example, a case where an image written as "'Tokyo ゛'" is input on a piece of paper will be explained with reference to Figure 2. do.

Character recognition 2 recognizes "To'", "Kyo", and "To" in this order, or, as shown in Figure 2, recognizes characters similar to "To" such as "To" and "Kyo". , “Kuruma natoga kyo” is “east”, “sadness”, “kyoya nato”,
For ゛都゜゛゛゛゛゛゛゛゛゛゛゛゛゛゛゛゛゛゛゛゛゛゛゛゛゛゛゛゛゛゛゛゛゛゛゛゛) As a result, “Kuma”, “Ai”, “Kyo”, “Part”, “Wholesale”, and “Gun” are excluded from the recognition target characters, so “To” is the candidate character 3. For ``゜゜東'' and ``京゜′, and for ``kyo'', ゛′kyo''′ and ``east''′, and for ``to'', ``to''.
' is passed to the word matching unit 4 as a valid candidate (FIG. 2(b)). Therefore, the combinations that can be considered as Chinese words are narrowed down to four: ゛゜Higashi-Toto'', ``Tokyo'''', ``Kyoto'''', and ``Kyo-Kyoto''. It is possible to determine that "'Tokyo"' is a valid candidate carved word. i;J. This is an explanation regarding the effect of the limited function of the character type.

Next, a specific example of the character type extraction section 8 will be explained. ,
FIG. 3 is a schematic diagram of an example of the character type storage section 5, the word recording section 6, and the character type extraction section 8. 1}1- In the word storage unit 6, prefecture names, city names, personal names, and other arbitrary single-word groups are registered. The character type storage section 5 has an area for storing a character type limited table consisting of character type set 1- corresponding to each word group. The operation of the character type extractor 8 is to extract ``゜Tokyo'' and ``Kanagawa Prefecture'' from the Tongo group of prefecture names.
``East'', ``Kyo'', ``To'' used in the word ``nato''
Character types such as ``,''``Kami'',``゛NA'', ゛NO1BI, and ゛゜ken are picked up and registered in the character type limited table of the prefecture name type set in the character type storage section 5. Similarly, character types are picked up from city names, people's names, and others, and registered as a corresponding character type set.

In character recognition, when recognizing a prefecture name, recognition is limited to the character types registered in the character type restriction table in the character type set 11 of the prefecture name in the character type storage unit 5. Similarly, for city names, personal names, etc., recognition is limited to character types that are registered in the character type set.

〔Effect of the invention〕

As explained above, the present invention extracts all character types from a word group and provides a character type extraction storage unit that stores them in the character type storage unit, thereby creating a limited character type table to be recognized. This is done automatically, which has the effect of eliminating the need for manual registration work in advance.

[Brief explanation of the drawing]

FIG. 1 is a block diagram of the configuration of a character reading device according to an embodiment of the present invention, and FIG. ) shows candidate characters when recognizing all character types, and FIG. 2 (1)) shows candidate characters when recognizing limited character types. FIG. 3 is an explanatory diagram of the operation of the character type extraction section 8 shown in FIG. 1. ]...Character image, 2...Character edict, Shikibu, 3.
... Candidate character, 4... Word matching section, 5... Character type storage section, 6... Immediate word storage section, 7 Candidate 1 Hayashi language, 8
Character type extraction part.

Claims (1)

[Claims] A word storage unit that stores a group of target words in advance, a character type storage unit that stores character types to be recognized in advance, and a character type that recognizes an input character image and stores it in the character type storage unit. a character recognition unit that outputs candidate characters limited to , and a word group in which words formed by combining candidate characters output from the character recognition unit for a plurality of the character images are stored in the word storage unit; In a character recognition device including a word matching unit having a function of checking whether a word exists in a word group, the character recognition device picks up all character types used in a word group registered in the word storage unit and stores them in the character type storage unit. A character recognition device comprising a character type extraction section.