JPS60204090A - Pattern recognition system and storage medium used in it - Google Patents

Abstract

PURPOSE:To improve the discrimination precision, the processing speed, and the economization of a device by storing individual reference pattern data for comparison and collation in a storage medium which can be replaced in individual information units. CONSTITUTION:A form 3b for recognition where characters to be read are hand- written is set to a prescribed position. A processing part 4 confirms that a storage medium M of a pertinent person is set to a prescribed position. Personal discrimination data corresponding to a specific person where characters are hand-written is inputted from an operation part 1; and if the processing part 4 confirms coincidence between this data and personal discrimination data stored preliminarily in the storage medium M, the processing part 4 outputs a signal, which writes character form comparison collation data D2 in a temporary storage part 5 of character form comparison collation data, to a data input/output part 8. The processing part 4 compares and collates a character pattern signal with character form comparison collation data D2 by a prescribed character comparison collation method to recognize a character.

Description

[Detailed Description of the Invention] [Technical Field of the Invention] The present invention collates input pattern information with pattern information from a storage medium in which a dictionary relating to hands or sounds of individual information units such as individuals is stored. The present invention relates to a pattern recognition method for performing recognition and a storage medium used therefor.

Technical costs of the invention and its problems 1 Conventionally, character recognition has been used as a method for optically or magnetically reading characters handwritten on documents such as OCR cards and performing predetermined recognition. ', rA 'B The standard glyph comparison and matching data (which is data that has extracted the features necessary for character comparison and matching from a large number of characters in advance) is stored in the main body at 'B'.
A method is known in which a plurality of sets of standard glyph comparison data (dictionaries) are set up and the read unknown characters are compared and recognized with this standard glyph comparison data.

In order to efficiently perform character recognition in such a character recognition system, it is necessary to unify the shapes of characters to be recognized as much as possible.

By the way, there are known methods for comparing and contrasting standard characters and unknown characters when performing character recognition, such as the pattern matching method, fixed point sampling method, contour feature extraction method, and composite similarity method. Regardless of which method is used to identify characters, it is difficult to identify handwritten characters if their shapes are far from standard characters. Therefore, it was necessary to impose considerable restrictions on the appearance of characters to be recognized.

In order to improve the recognition accuracy in such conventional character recognition systems, for example, the number and number of sets of standard character form comparison data to be set in the character recognition device must be increased. In other words, it is necessary to input human handwriting of at least several hundred angstroms. Therefore, when trying to ensure a certain level of recognition accuracy for a variety of handwritten characters,
A problem arises in that the cost of the device increases due to the creation of standard character form comparison and verification data and the need for a large-capacity storage device within the device.

Furthermore, as the number and number of phases of standard character form comparison data are increased, a problem arises in that the processing speed decreases. There is a method described in Japanese Unexamined Patent Application Publication No. 84372/1983 that addresses these points. However, the capacity of the standard dictionary is still large, and the personal dictionary is only used for correction.

However, after considering processing speed, equipment cost, etc.
The reality was that a compromise had to be found regarding the degree of regulation of the writing style of characters to be recognized, the number of standard character form comparison data, and compositional problems.

In addition, the character recognition device that realizes the conventional character recognition method is
It is constructed on the premise that characters can be recognized by an unspecified number of scribes, but in normal office work, the number of scribes is limited to a specific group of people, and the amount of data in the dictionary used by individuals is small. There was a problem that it became uneconomical.

Moreover, it had drawbacks such as being a major obstacle to realizing portability.

[Object of the Invention] The present invention has been made in view of the above-mentioned conventional circumstances, and by replacing pattern signals such as characters and sounds with dictionaries for each individual information unit, it is possible to improve discrimination, processing speed, and so on. The object of the present invention is to provide a pattern recognition method that improves the economical efficiency of a rice cracker and a storage medium used therefor.

[Summary of the Invention] In other words, the pattern recognition method of the present invention stores comparison and matching data in units of individual information, such as individuals, from which features necessary for comparison and matching of characters, sounds, etc. are extracted, in an exchangeable storage medium. The method is characterized in that when the pattern is recognized, the data is read from the storage medium storing the reference pattern data for comparison and verification of the individual information unit and used as a recognition standard. The storage medium used is
It is characterized in that reference pattern data for comparison and verification, in which features necessary for comparison and verification of individual information are extracted, is stored in advance.

In conventional character recognition methods, standard character comparison comparison data and/or individual reference pattern data for comparison and verification, which are reference pattern data for comparison and verification during data processing, are stored in advance in a storage unit within the main body of the device. However, in the method of the present invention, individual reference pattern data for comparison and verification is stored in a storage medium that can be exchanged in units of individual information.
When input information is recognized, the corresponding individual comparison and verification data is read out from this storage medium and used as comparison and verification data.

[Embodiment of the Invention] Hereinafter, details of the present invention will be described with reference to an embodiment shown in the drawings.

FIG. 1 is a block diagram for explaining an embodiment in which the method of the present invention is applied to a character recognition device.

In the figure, reference numeral 1 is an operation unit that is equipped with a key switch operated by an operator and generates an instruction signal according to the operation, and reference numeral 2 is an operation unit that is equipped with a key switch that is operated by an operator, and 2 is a control unit that generates an instruction signal that is recorded on a registration form 3a such as an OCR card and a recognition form 3b. A reading unit that optically or magnetically reads handwritten characters; 4 a processing unit that performs various data processing in character recognition; 5 a provisional storage unit for character shape comparison and collation data that will be described later; 6 a data synthesis unit that will also be described later; and 7 a recognition unit. A display unit such as a cathode ray tube display device that displays the results;
8 is writing of reference pattern information, such as personal identification data D1 and character shape comparison data D2, which will be described later in 1-, into an exchangeable storage medium M such as an IC card, a laser card, or a memory bag. and a data output unit for reading.

The storage medium M1 glyph comparison and verification data provisional storage section 5 and data synthesis section 6 will be described in detail below.

FIG. 2 is a conceptual diagram illustrating storage medium Mk: stored data. As described above, the storage medium M stores the personal identification data D1 and the character shape comparison verification data D2.

The personal identification data D1 is a code number or identification name corresponding to each person in a specific group consisting of people who handwrite characters to be recognized on the recognition form 3b.

The character shape comparison and verification data D2 is obtained under the condition that there are no special regulations on the writing style of each person identified by the individual data D1 (it is desirable to have a rule 11 such as "write carefully").
This is a data group in which predetermined features are extracted from a group of handwritten characters, and is composed of character identification codes and corresponding feature data.

In addition, personal identification data D1 and character shape comparison verification data D
2 is actually digitally encoded and stored.

The data synthesis section 6 is a section that synthesizes the personal identification data D1 and the character shape comparison and verification data D2, and adds, for example, the personal identification data D1 to the head portion of the character shape comparison and verification data D2.

Next, personal identification data D1 and character shape comparison verification data D2
The process of registering the data in the storage medium M will be explained step by step.

(1) An individual who handwrites characters that should be recognized later,
A registration form 3a handwritten with characters that are scheduled to be used without any special restrictions on how they are written is set in a predetermined position of the device.

(2) When the "system activation" operation is performed on the operation unit 1, a signal for activating the system is output to each part of the device.

(3) When the operation section 1 performs the operation of "registering the character shape comparison verification data D2 in the storage medium M," an instruction signal thereof is outputted to the processing section 4.

(4) When a code number or identification name, which is personal identification data D+, is entered in the operation section 1, the signal is output to the data synthesis section 6 and temporarily recorded.

(5) The processing section 1411 4 confirms that the storage medium M is set at a predetermined position.

(6) When the "Start Reading" operation is performed on the operation section 1, the instruction signal is output to the processing section 4, and the processing section 4 sends the characters handwritten on the registration form 3a to the reading section 2. Outputs the command to perform the reading.

(7) Every time the reading unit 2 reads a certain number of handwritten characters written on a form, a character pattern signal is output to the processing unit 4, and the processing unit 4 receives the character pattern signal from the character pattern signal.
The characteristics of each character are extracted as necessary information in any of the adopted character comparison methods such as the pattern matching method, fixed point sampling method, contour feature extraction method, or composite similarity method, and are stored in the registration form 3a. The corresponding character identification codes are combined and temporarily stored in the glyph comparison and collation data temporary storage section 5 as glyph comparison and collation data D2.

(This is repeated until Liebe's handwriting is read).

(8) The processing unit 4 reads the character shape comparison and verification data D2 from the character shape comparison and verification data temporary storage unit '5, and outputs it to the data synthesis unit 5 where the personal identification data D1 is temporarily stored.

/ (9) The data synthesis section 6 synthesizes the personal identification data D1 and the individual character shape comparison verification data D2, and outputs the synthesized data signal to the data input/output section 8.

(10) The data input/output unit 8 writes the data signal synthesized in (9) onto the storage medium M.

(11) Perform professional wrestling for each person constituting a specific group for more than the end, and create - sets of exchangeable storage media M.

In addition, as the registration form 3a, for example, as shown in FIG. Print it out, provide a corresponding blank side 301 for each, and have a specific individual use the recognition form 3b.
If you use a device in which characters that may be handwritten are placed in a predetermined blank space, and the corresponding character identification code is stored in the character synthesis unit 6, it will be added to the registration form 3a each time. By adding necessary characters and adding the corresponding character identification code to the data synthesis section 6, the number of data in the individual comparison verification data D2 can be increased.

In this case, the storage medium M can be replaced according to changes in content by selecting a medium on which information can be rewritten.

Next, a process will be described in which characters handwritten by a specific individual on the recognition form 3b are recognized by the above device.

(1) When the "system activation" operation is performed on the operation unit 1, a signal for activating the system is output to each part of the device.

(2) Set the recognition form 3b on which the characters to be read are handwritten in a predetermined position.

(3) When a "character recognition" operation is performed on the operation section 1, an instruction signal thereof is outputted to the processing section 4.

(4) The processing unit 4 confirms that the storage medium M of the individual is set in a predetermined position.

(5) Enter the personal identification data corresponding to the specific individual who handwritten the characters from the operation unit 1, and the processing unit 4 writes the characters on the medium M
When the processing unit 4 confirms the match with the personal identification data D1 stored in advance in -C, the processing unit 4 sends the glyph comparison matching data D2 to the data input/output unit 8, and transfers the glyph comparison matching data D2 to the glyph comparison matching data provisional storage unit 5.
Outputs a signal to write to.

(6) When the "Start character recognition" operation is performed on the operation section 1, the instruction signal is output to the processing section 4. 1
The first unit 4 outputs a signal to the reading unit 2 to read the characters handwritten on the recognition form 3b.

(7) Every time the reading unit 2 reads a certain number of handwritten characters, a character pattern signal is output to the processing unit 4 (this is repeated until all handwritten characters have been read)
.

(8) Character pattern signal and character shape comparison comparison data written in the processing unit 4 in the character shape comparison window 0 data D
2 by a predetermined character comparison method, and character recognition is performed.

(9) As a result of the recognition, the processing unit 4 outputs to the display unit 7 a signal that causes the standard form of the recognized character, the code number corresponding to the recognized character, etc. to be displayed.

(10) If there is an error in the recognition result displayed on the display section 7 or there are unrecognizable characters, perform a correction operation on the operation section 1.

(11) Finished.

In the above professional wrestling, characters handwritten on the recognition form 3b are recognized for each specific individual.

It should be noted that the method of the present invention is not limited to the above-mentioned embodiment, but standard character form comparison and matching data similar to conventional character recognition methods is stored in the main body of the device, and recognition standards are set as necessary. It may be possible to do this.

As the storage medium M of the present invention, an IC card for mobile phones 91), which has been increasingly popularized in recent years, is suitable.

An IC card is a conventional card with a magnetic stripe (
For example, it is attracting attention as an alternative to credit cards, bank cards, etc., and as shown in FIG.
It is equipped with a memory depth 10 such as SI or VLSI and a microprocessor chip 11, and has a much higher information storage capacity than conventional magnetic cards.

Therefore, other information such as individual voiceprint information, a dictionary for word processors in a specific range, a dictionary of individual keyword systems for searching document file devices, etc. can also be stored together.

Furthermore, in the future, it is thought that a large amount of individual information will be stored on a single IC card, but it is recommended that the aforementioned individual character shape comparison data be stored in advance as personal information. For example, according to the method of the present invention,
It will be possible to have devices recognize handwritten characters in various places. Therefore, the above-mentioned specific groups can be extended to the national level and can be applied to a wide range of applications.

In the above embodiment, an example was explained in which the personal identification data D1 was input to the storage medium, but the personal identification data D
1 is not necessarily necessary. [Effects of the Invention] As explained above, in the method of the present invention, reference patterns are stored in advance in units of individual information in a set of exchangeable storage media, and when recognizing patterns, Since pattern information to be recognized, for example, character shape comparison data of the same person who handwritten the characters, is used as successive comparison data, there is a significant effect on recognition discrimination accuracy, processing speed, and economical efficiency of miniaturization of the device.

[Brief explanation of the drawing]

FIG. 1 is a block diagram showing a system configuration for explaining an embodiment in which the method of the present invention is applied to a character recognition device, and FIG. 2 is a conceptual diagram explaining data stored in the storage medium M of the present invention. , FIG. 3 is a plan view showing a registration form used in a character recognition device to which the method of the present invention is applied, and FIG. 4 is a plan view showing an IC card as an embodiment of the storage medium M of the present invention. . 1.........Operation unit 2...Reading unit 3a...Registration form 3b... ...Recognition form 4...Processing unit 5...Glyph comparison verification data temporary storage unit 6... ...Data synthesis section 7...
......Display section D1...Personal identification data 'D2...
...Individual glyph comparison verification data M...
・・・・・・Patent attorney for storage media Satoshi Suyama - Figure 1 Figure 2

Claims (8)

[Claims] (1) When performing recognition by comparing reference pattern information stored in a storage medium and cover pattern information of a person to be recognized in the processing main body, the reference pattern information is stored in individual information units, and the processing main body A pattern recognition method characterized in that a storage medium configured to be exchangeable for a pattern is selected and recognized according to the contents of the cover pattern information. (2) The pattern recognition method according to claim 1, wherein the pattern information is character pattern information. (3) The pattern recognition method according to claim 1, wherein the pattern information is voice pattern information. (4) The pattern recognition method according to claim 1, wherein the individual information unit is information on an individual basis. (5) The pattern recognition method according to claim 1, wherein the individual information unit is information for each technical field. (6) The character recognition system according to claim 1, wherein the storage medium is a plurality of RC cards each having integrated circuit chips mounted inside a resin plate-like body. (7) A storage medium that stores unique reference data for comparison and verification in which features necessary for comparison and verification in units of information in a specific individual field are extracted, and is configured to be detachable. (8) The storage medium according to claim 7, which is constituted by an FC card having an integrated circuit chip mounted inside a plate-shaped body made of synthetic resin.