JPH0253099A - Character recognition device - Google Patents

Abstract

PURPOSE:To recognize voices and handwritten characters of many persons by reading a standard feature pattern extracted from an input handwritten character image by using a portable storage medium at the time of character recognition. CONSTITUTION:This device is provided with a writing means 300 which writes the standard feature pattern extracted by an extracting means 100 from the voice of a person to be recognized on the portable storage medium 200 assigned to the person to be recognized and a reading means 400 which reads the standard feature pattern out of the portable storage means 200 so as to identify the character of an input voice. Thus, the standard feature pattern is written on the portable storage medium 200 and when the character recognition is performed, the standard feature pattern is read out of the portable storage medium 200. Consequently, the voices or handwritten characters of many persons can be recognized and the character recognition device main body does not require any memory for storing the standard feature pattern.

Description

DETAILED DESCRIPTION OF THE INVENTION [Field of Industrial Application] The present invention relates to a character recognition device that recognizes character information input in the form of voice or image information and converts it into a character code.

[Prior technology] At present, artificial intelligence technology using microcomputers is applied to character recognition devices that input sentences using human voice through a microphone, and character recognition devices that input sentences by hand from a coordinate input lid called a tablet. Character recognition devices have been developed that input the meaning of sentences handwritten on paper using a facsimile or an image scanner.

Conventional character recognition devices of this type convert character information input in the form of audio or image patterns into a character pattern that represents its characteristics, and then compare it with a standard pattern stored in memory in advance. There is. As a result of this comparison, a standard pattern character code that matches the input character pattern is output as the identification result of the input character information.

For example, in speech recognition, the above-mentioned standard pattern is characterized by the input voice information being input by the user into a microphone sequentially such as "a", "i", and "u" in advance, and then inputted by the character recognition device. The pattern, specifically the power spectrum for each audio frequency, is converted into a character code “A′
Characteristic patterns corresponding to ``,'', ``i'', ``u,'', etc. are stored in the internal memory as standard patterns.

[Problems to be Solved by the Invention] However, conventional character recognition devices of this type have the disadvantage that during speech recognition, if a person different from the person who registered the speech for the standard pattern inputs the speech, erroneous recognition often occurs. . The same thing can also be said for character recognition of image patterns.

In order to overcome these drawbacks, it may be possible to register characteristic patterns of multiple people as standard patterns in advance in the character recognition device, but for this purpose, it is necessary to store the memory capacity of the standard patterns to be matched. It is expected that another problem will arise in that the character recognition device will take a long time.

SUMMARY OF THE INVENTION An object of the present invention is to eliminate the above-mentioned problems and to provide a small character recognition device that is capable of recognizing character information in the form of speech or images input by a large number of people.

[Means for Solving the Problem] In order to achieve such an object, the first embodiment of the present invention extracts a feature pattern from input speech, and converts the extracted feature pattern into a pattern whose character type is known in advance. A character recognition device that identifies the type of character indicated by an input voice by comparing it with a standard feature pattern, the character recognition device comprising: an extraction means for extracting a standard feature pattern from the voice of a person to be recognized; The present invention is characterized by comprising a writing means for writing the standard feature pattern into a portable storage medium assigned to a person to be recognized, and a reading means for reading out the standard feature pattern from the portable storage medium for identifying characters of input speech. shall be.

The second form of the present invention extracts a feature pattern from an input handwritten character image, and compares the extracted feature pattern with a standard feature pattern whose type of character is known in advance. A character recognition device for identifying character types, comprising an extraction means for extracting a standard feature pattern from an image of handwritten characters of a person to be recognized, and a portable character recognition device in which the standard feature pattern extracted from the extraction means is assigned to the person to be recognized. The present invention is characterized by comprising a writing means for writing into a storage medium, and a reading means for reading out a standard feature pattern from a portable storage medium for character identification of an input handwritten character image.

[Function] In the first embodiment of the present invention, a standard feature pattern is extracted from the input voice of the person to be recognized, and in the second embodiment of the present invention, from the input handwritten character image by an extraction means, and the standard feature pattern is stored in a portable storage medium. By writing a pattern and reading the standard feature pattern from a portable storage medium during character recognition, by assigning a portable storage medium to each person to be recognized, it is possible to recognize speech or handwritten characters from many people. becomes possible. Further, the character recognition device main body does not require a memory for storing standard feature patterns, and is smaller and requires shorter recognition processing time than a conventional character recognition device equipped with a memory for storing standard feature patterns for a large number of users.

[Example] Hereinafter, an example of the present invention will be described in detail with reference to the drawings. In this example, pattern recognition of handwritten character images will be described.

FIG. 1 shows the basic configuration of an embodiment of the present invention.

In FIG. 1, a dash-dotted line block 1000 extracts a feature pattern from an input handwritten character image, and compares the extracted feature pattern with a standard feature pattern whose type of character is known in advance. 1 shows a character recognition device that identifies the type of character indicated by a handwritten character image.

Reference numeral 100 denotes an extraction means for extracting the standard feature pattern from a handwritten character image of a person to be recognized.

200 is a portable storage medium.

A writing means 300 writes the standard feature pattern extracted from the extraction means 100 into a portable storage medium assigned to the person to be recognized.

Reference numeral 400 denotes a reading unit for reading out the standard feature pattern from the portable storage medium for identifying characters in the input handwritten character image. Note that in the case of voice recognition, the input handwritten character image becomes the input voice.

FIG. 2 shows the appearance of an IC card as an example of a portable information storage medium.

Since the structure of the IC card itself is generally known, only the points related to the present invention will be explained.

In FIG. 2, an IC card 1 houses an electrically rewritable memory device (E2FROM) and a semiconductor chip 2 incorporating a processor capable of various arithmetic functions such as reading and writing information to and from the E2FROM. ing. E2FROM combines the identification information assigned to each individual, a dictionary for character type determination that uses as a standard pattern a characteristic pattern extracted from the image pattern of registered characters, for example, the Saigen font, and this dictionary for character identification. Characteristic information of characters that differ depending on each person using the system is stored.

The lead terminal 3 is a terminal for inputting and outputting signals to and from the semiconductor chip 2, and is provided on the surface of the IC card 1.

FIG. 3 shows a schematic configuration of a character recognition device that inputs information to the IC cart 1, reads out the information, and performs character recognition processing.

In FIG. 3, 10 is a host computer generally called a personal computer unit.

The host computer 10 includes an information processing device body 10-3 having a central processing unit (CPU) as an extraction means;
It consists of a keyboard 10-1%, a CRT display 1 row 2, a facsimile 10-4, and a floppy disk storage device (not shown).

This floppy disk is used to store programs and the like for performing character recognition processing.

Reference numeral 11 denotes a card reader/writer which reads/writes information to/from the IC cart 1 as reading means and writing means.

FIG. 4 shows a specific circuit configuration of the host computer IO shown in FIG.

In FIG. 4, keys 1O-IA provided on the keyboard 10-1 are keys for instructing reading of information from the IC card 1. Keys 1O-IB are keys for instructing input of handwritten characters. In addition, these keys 1O-IA, 1O-
Instead of the IB, the above-mentioned instruction may be given in the form of a control command (command code) from the keyboard 10-1.

In the information processing device main body 1O-3, 1O-3A is the fifth
A central processing unit (CPU) executes the control procedure shown in the figure.
). 10-38 is a floppy disk storage device (
FDD).

1O-3G is a read-only mailbox that stores control procedures for system operation in addition to the control procedures shown in FIG.
J (TIOM) Yes. to-30 is CPU1o-
This is a random access memory (RAM) that stores various information used when executing the 38 power control procedure. key 1
When CPIJlo-:IA detects the instruction from 0-18, CPJllO-3^ stores the storage information of IC cart 1 in FDDIO-:ID via card reader/writer 11.

FIG. 5 shows the control procedure executed by CPt11O-3A of this embodiment.

In Figure 5, first, write your own name in handwriting such as 'R'.
In the case of El, the user inserts the IC card 1 into the card reader/writer 11 and sets the paper on which his/her name is written into the facsimile 1O-4. Next, keyboard 1
The facsimile machine 0-1 instructs the facsimile machine 10-1 to register its own name and instructs the facsimile machine 10-4 to read and scan. This instruction is detected by CPLIIO-3A and
0-3A extracts character characteristic patterns from the character information read by the facsimile according to a well-known control procedure (step 5o-510).

Next, CPIJIO-3A was extracted from among the standard feature patterns written in the dictionary of the IC card 1. Search for the standard feature pattern that is most similar to the feature pattern of the Gigi characters.

Next, CPIIIO-3A displays the character coat or character image corresponding to the standard feature pattern found as a result of this search on RTIO-2, and instructs RTIO-2 whether or not the character image input by the user is correctly recognized. request the user. If the recognition result is correct, an instruction to discard the CPLIIO-3 is received from the keyboard 10-1, and if the recognition result is incorrect, the character code to be correctly recognized is received from the keyboard 10-1.

Next, CPU10-3A calculates the difference between the characteristic pattern of the input character image and the standard characteristic pattern of the dictionary corresponding to this characteristic pattern, and uses the calculation result as the characteristic of each individual's writing style. Write to cart 1 (step SIO→512→515).

Characteristics of writing patterns include, for example, the inclination angle of horizontal lines and vertical lines, the presence or absence of splashes at the end points of lines, etc. It is preferable to store these extracted features in the memory area IJB of the IC card 1 (see FIG. 6).

Next, the CPU 10-3A stores the identification code input from the keyboard 10-1 in the memory area MA, and completes the registration of the characteristic pattern of handwriting for each individual.

In this embodiment, in order to recognize various characters, a dictionary that describes standard feature patterns and their character coats associated with each character code, and feature patterns extracted from each individual's unique habits are used for each character code. Although it is stored in IC card 1 as a standard characteristic pattern for individuals, if there are few characters to be recognized, for example, if you use IC card 1 as a credit card and want to use the input characters instead of a PIN number, , it is possible to use the characteristic pattern itself of the input handwritten character image of each individual as the standard pattern.

Next, the IC card l whose standard feature pattern has been registered is
We will explain the procedure for character recognition using .

When the user instructs character recognition to be performed from the keyboard 10-1 and sets paper to be recognized in the facsimile machine 10-4, the CPt11O-3A reads the first standard character pattern in the dictionary and each individual's writing style. By comparing the standard pattern created based on the second feature pattern and the feature pattern extracted from the input information, the recognized Suhegi characters are compared (steps 51 to S5).

Next, the character code recognized as a result of this verification is visually displayed to await confirmation and correction by the user. If there is a correction, the characteristic pattern of the character code to be corrected as instructed by the user is compared with the characteristic pattern of the input character information, the difference is newly extracted as a writing pattern, and the character characteristics of the individual on IC card 1 are extracted. The pattern is rewritten and this control procedure ends (steps 56 to S8-3).

In this way, the IC card stores character patterns unique to each individual's characters, which improves the character recognition rate for each individual's characters, and also improves the ability of the device to recognize characters from a large number of people. There is no need to make the device itself larger than before.

Further, in the case of this embodiment, character recognition of character images has been described, but in the case of voice character recognition, the following may be performed.

That is, a first standard feature pattern extracted from a specific person (corresponding to the dictionary of this embodiment) is stored in the IC card l, and this standard feature pattern and a feature pattern extracted from the voice of the person to be recognized at the time of registration are stored in the IC card l. Specifically, the deviation in the power spectrum in each frequency band (corresponding to the writing pattern in this embodiment) may be stored as a characteristic pattern of each individual's voice.

[Effects of the Invention] As described above, according to the present invention, standard feature patterns are extracted by the extraction means from the input voice of the person to be recognized in the first form and from the input handwritten character image in the second form. By writing this standard feature pattern into a portable storage medium and reading it out from the portable storage medium during character recognition, by allocating a portable storage medium to each person to be recognized, a large number of It becomes possible to recognize human voice or handwritten characters. Further, the character recognition device main body does not require a memory for storing standard feature patterns, and is smaller and requires shorter recognition processing time than a conventional character recognition device equipped with a memory for storing standard feature patterns for a large number of users.

[Brief explanation of the drawing]

FIG. 1 is a block diagram showing the basic configuration of an embodiment of the present invention. FIG. 2 is a sectional view showing a schematic configuration of an RC card according to an embodiment of the present invention. FIG. 3 is a side view showing the external configuration of an embodiment of the present invention. , FIG. 4 is a block diagram showing the circuit configuration of the host computer IO shown in FIG. 3, FIG. 5 is a flowchart showing the control procedure executed by the host computer IO shown in FIG. 3, and FIG. 6 is an embodiment of the present invention. This is a memory map showing the memory configuration of the IC card. IO...Host computer, 10-1...Keyboard, 1O-2...CRT. 1O-3A-CPII. 10-4...Facsimile, II...Cart reader/writer. 24 Kyou ramie 27゜10 ding, Nut:・=°Uni 10-1 ん-1-One heather so 1 bunch fern u) IJj, (Kenmen drawing piece 7 0
'-J

Claims (1)

[Claims] 1) By extracting a feature pattern from the input voice and comparing the extracted feature pattern with a standard feature pattern whose type of character is known in advance, the character indicated by the input voice is determined. A character recognition device for identifying a type, comprising: an extraction means for extracting the standard feature pattern from the voice of a person to be recognized; and a mobile phone assigned to the person to be recognized, the standard feature pattern extracted from the extraction means. A character recognition device comprising: a writing device for writing into a storage medium; and a reading device for reading out the standard feature pattern from the portable storage medium for identifying characters of the input voice. 2) The type of character indicated by the input handwritten character image is determined by extracting a feature pattern from the input handwritten character image and comparing the extracted feature pattern with a standard feature pattern whose character type is known in advance. A character recognition device for identifying a character, comprising an extraction means for extracting the standard feature pattern from a handwritten character image of a person to be recognized, and a portable character recognition device that assigns the standard feature pattern extracted from the extraction means to the person to be recognized. A character recognition device comprising: a writing device for writing into a storage medium; and a reading device for reading out the standard feature pattern from the portable storage medium for identifying characters in the input handwritten character image.