JPS63100585A - Character recognizing device - Google Patents

Abstract

PURPOSE:To easily vary a read character kind in accordance with a read object, by recording a recognition dictionary added newly by learning, independently of a recognition dictionary prepared in advance. CONSTITUTION:When a reject character is generated in recognizing a character, a screen as shown in the figure is displayed on a CRT 9. In this case, when an instruction to execute learning is given from a keyboard 10, a character learning part 11 receives a character pattern from a character recognizing part 5 by a control part 13, and generates a recognition dictionary to its character. The generated recognition dictionary is loaded to a dictionary memory 12, and also, written in a disk 8, as well. The dictionary prepared in advance, and the learned dictionary are housed in a hard disk 7 and a floppy disk 8, respectively. In this way, in accordance with the field and property of an object document, separate floppy disks can be prepared, and by using a group of dictionaries being suitable for the object document, the recognition can be executed.

Description

[Detailed description of the invention] [Purpose of the invention] (Industrial (1) field) This invention relates to character recognition installation.

(Conventional technology) Conventional character recognition Minato! When it comes to smell, recognition dictionaries can only recognize characters that have been prepared in advance.

When a document to be recognized contains unknown symbols or different fonts, it is not possible to deal with it on the spot, and a new recognition dictionary for the characters must be created offline.

Therefore, in recent years, we have added a learning function to the single character recognition system 1, making it possible to create a T:E1 recognition dictionary for unknown symbols and unknown fonts in the field where reading work is performed. began to be considered.

(The gap that the invention attempts to solve) A sentence recognition device with such a learning function.

The recognition dictionary for the learned characters is added 710 to the recognition dictionary 11I# prepared in advance.

However, the unknown symbols and unknown fonts that you should learn are usually different for each book, document, and form.
When reading a document other than the one for which learning has been performed, there is a possibility that unnecessary processing will be required or erroneous recognition will occur because of the learning.

In addition, if the number of sentences that can be learned is limited due to system limitations, unknown symbols in the current document to be read can be learned because they were learned for another document. It is possible that this will not be possible.

[M base of invention]

(Means for Solving the Problems) In order to solve the problems in the present invention, the following measures are taken.

In other words, a recognition dictionary created through learning.

Instead of recording by adding to a recognition dictionary prepared in advance, record the learned dictionary and the dictionary prepared in advance separately, and when loading the recognition dictionary, merge and load the two. .

(Function) With this means, it is possible to have a separate learning dictionary for each type of object, and it is possible to use as many learning dictionaries as necessary.

(Implementation?'lJ) A line fishing embodiment of the present invention will be explained below with reference to the drawings.

FIG. 1 shows the configuration of a single character recognition device.

The form 1 input directly into the book is read by a scanner 2 and stored in a buffer 3.

From one batch of data stored in buffer 3.

Each character is cut out by the character cutting section 4.

It was cut out! ! ! Each character is sent to the 1-character recognition Pi section 5.The character recognition section recognizes each character using the recognition word "a1!-" loaded into the dictionary memory.

Output the result. However, the dictionary used in single character recognition 81S5 is
When starting up the system, the dictionary loading section 6 loads the dictionary from the floppy disk 8 across the eight-dimensional disk 7.

If a reject character occurs during single character recognition, a screen like that shown in FIG. 2 is displayed on the CRT 9j. At this time, when an instruction t- to perform learning is given from the keyboard lO, the character learning g is performed by the control item 13.
ll receives the character pattern from the single character recognition unit 5,
Create &1a dictionary for that character.

The created recognition dictionary is loaded into the dictionary memory 12 and also written to the disk 8.

In the system as described above, the word wh, which is prepared in advance, is learned in the head disk 7, and the word VII is
I decided to store it on floppy disk 8.

By doing this, it is possible to prepare separate floppies depending on the field characteristics of the target document, and the present invention can be performed at once by using a set of dictionaries suitable for the target document. is not limited to the above-mentioned actual tIIi example.

Various modifications are possible without changing the gist.

If you line up, the learning word *t-, S instead of floppy disk.
Record to a separate file on the same eight-dimensional disk.

In addition, single character recognition is used for voice recognition.

【Effect of the invention〕

According to the invention, therefore, depending on the object to be read.

It is possible to easily change the 9-character box.

[Brief explanation of the drawing]

FIG. 1 is a diagram showing the configuration of a character recognition tI & device, and the second diagram shows a learning method for rejected characters. 5...Character w1 identification part, 6...Dictionary loading part, 7...
・Eight disk, 8...Floppy disk, 11・
...Character learning section, 12...Dictionary memory. Agent Patent Attorney Ken Nori Chika Udo Kikuo Takehana

Claims (1)

[Claims] A character recognition device that can appropriately learn and recognize characters other than the character types prepared in advance in the recognition dictionary is characterized by recording newly added recognition dictionaries through learning separately from the recognition dictionaries prepared in advance. character recognition device.