JPH03268089A - Optical character reader - Google Patents

Abstract

PURPOSE:To omit the troublesome correction of characters by recognizing the characters based on the subset information including a corrected character after an operator corrects the character which is not included in the subset information. CONSTITUTION:A photoelectric transducer part 1 is provided together with a character recognizing part 2, a subset information storage part 3, a reading control part 4, an operating part 5, and a display part 6. When the recognizing result is corrected, a subset information control means 41 functions to reflect the result of correction onto the subset information stored in the part 3. Hereafter a character range is limited by the subset information and therefore the characters are recognized in response to the change of a subset range. Thus each troublesome correction of characters can be omitted.

Description

DETAILED DESCRIPTION OF THE INVENTION [Object of the Invention] (Field of Industrial Application) The present invention relates to an optical character reader (OCR).

(Conventional technology) Conventionally, in OCR, in order to improve reading accuracy,
A method is used to recognize characters by limiting the types of characters to be read for each character entry field. For example, if you know in advance that only numbers will be entered, such as in an amount column,
``Number'' is specified as the character to be read, and in the case of a column where only katakana characters are entered, such as a furigana column, ``katakana'' is specified as the character to be read. Similarly, in the case of fields in which kanji are entered, such as name fields and address fields, kanji should be specified as the characters to be read, but in the case of kanji, there are many types of characters (JIS- In the first level alone, there are over 2,000 character types), even for the same kanji, in the name field, only the character types that are often used in names are specified as the characters to be read, and in the address field, only the character types that are often used in addresses are specified as the characters to be read. specify. Such a designation is referred to as a subset designation or simply a subset, and information on characters within this designated range is referred to as subset information.

(Problems to be Solved by the Invention) However, in such conventional OCR, when a subset is specified, if there are characters that are not included in the subset information, the character may be recognized incorrectly or may not be recognized. For example, when reading a form such as a bank transfer request form, fields such as the bank name and branch name of the transfer destination are read using the bank name subset and branch name subset, respectively. When a branch is established (X), if this bank name or branch name is not included in the subset, it will be misread or unrecognizable.

Therefore, in such a case, an operator must intervene to correct the results, resulting in a problem that the process is complicated.

The present invention was made to solve the problems of conventional OCR, and its purpose is to prevent characters from appearing in the subset information and being modified by the operator. An object of the present invention is to provide an optical character reading device that can perform character recognition using included subset information.

[Structure of the Invention] (Means for Solving the Problems) The present invention comprises: a photoelectric conversion means for charging and converting a form image to obtain an image signal; a recognition dictionary unit storing various parameters used during character recognition; a subset information storage means storing subset information for limiting the range of characters related to the character range; and when necessary, the character range is limited by the subset information stored in the subset information storage means and the parameters of the recognition dictionary section are referred to. , a character recognition means for performing character recognition based on the image signal obtained by the photoelectric conversion means, a modification means for modifying the recognition result of the character recognition means, and a subset in the subset information storage means that performs the modification by the modification means. The optical character reading device is provided with a subset information management means that performs processing to reflect the information.

(Operation) According to the above configuration, when a recognition result is modified, the subset information management means operates, and the modification is reflected in the subset information in the subset information storage means, so that when it is desired to add a character range, etc. This means that it can be done.

(Example) Hereinafter, an example of the present invention will be described with reference to the drawings. FIG. 1 is a block diagram of one embodiment of the present invention. In the figure, 1 indicates a photoelectric conversion unit, which includes an image scanner such as a CCD, a noise removal circuit, an A/D conversion circuit, etc., and converts photoelectrically converted analog signals into digital image signals used for character recognition. and gives it to the character recognition section 2. The character recognition unit 2 includes a recognition dictionary unit 21 that stores parameters for similarity detection used for pattern matching methods, feature extraction methods, etc., and performs character recognition using the above-mentioned methods. conduct. 3 indicates a subset information storage unit;
Subset information that limits the range of characters related to recognition as described above is stored. When the character recognition unit 2 receives an instruction from the reading control unit 4, the character recognition unit 2 performs recognition by limiting a character range based on the relevant subset information in the subset information storage unit 3.

The reading control section 4 receives a subset instruction from an operation section 5 constituted by a keyboard input device, etc., and supplies it to the character recognition section 2. The result recognized by the character recognition section 2 is given as a character code to the reading control section 4. The reading control unit 4 provides the second character code to the display unit 6. The display section 6 includes, for example, a CRT, its controller, a pattern generator, etc., and patterns the character code given from the reading control section 4 to display corresponding characters. At this time, the operation section 5 functions as a correction means. That is, the operation unit is equipped with a correction key, a cursor key, a next candidate indication key, a character key, etc., and by positioning the cursor below the character to be corrected, operating the correction key, and then selecting the next candidate indication key or character key. Input the correct characters by operating the keys. The character code of this character is given to the display section 6 from the operation section 5 or from the character recognition section 2 via the reading control section 4 and displayed. When the correct character is displayed, move the cursor and continue making corrections in the same way. The corrected data is transferred to and held in a storage device (not shown) by operating a key on the operation unit 5 that instructs confirmation.

During a correction operation after character recognition using the above-mentioned subset information, the subset information management means 41 in the reading control section 4, which is comprised of a CPU, operates according to the flowchart shown in FIG. 2, for example. That is, it detects whether or not a correction input is made from the operation unit 5 (5IOI), and if so, controls the display as described above and holds the corrected content (character code) (5IOI). S102). Then, until the correction is finished, step 5IOI, 51
02 is repeated (S103). To explain a specific example here, it is assumed that only the bank name column is read in the form of a bank transfer request form as shown in FIG. At this time, a subset of the bank name is specified, but since "Toshiba Bank" is a new bank, this subset information includes "
It does not contain the character ``shiba'' and is misrecognized.

Therefore, the correction is made and the character ``Shiba yo'' is displayed on the display unit 6, and its character coat is held in the subset information management means 41.2 When only this character ``Shiba'' is corrected, the process ends. In this case, the subset information management means 41 receives an instruction (from the operation unit 5) to add the retained content (in this case, the character "shiba") to the subset information of the bank name.
A key operation) is detected (S104). If there was an instruction here, the reading mode would have been switched (i.e. whether to switch to reading by another subset) (5105), but if it had been switched, the relevant subset information (bank name subset information) 8106
), it is detected whether the content should be added by checking whether the characters of the retained content are included (S107). In the above example,
It is checked whether the character "Shiba" is included in the bank name subset, and if it is not included, it is added (8108). Then, it is checked whether there is any other correction content (S109), and if there is, the process returns to step S106 to continue the operation, and when there is no more processing related to the correction content, the process returns to the main program. In this way, a new range of characters is added to the subset information and used for next time recognition, eliminating erroneous recognition and
There is no need to perform correction operations due to unrecognizability. In this way, in this embodiment, the subset information management means 4
As shown in FIG. 1, determination means 410 determines whether or not the characters resulting from the correction should be added to the subset information (for example, steps 5104 and 5105), and the determination result of this determination means 41. It consists of a rewriting means 420 that rewrites the subset information in the subset information storage means in accordance with the information stored in the subset information storage means.

However, if you want to always automatically incorporate modifications into the subset information, step.

5104 is unnecessary. In addition, if the load of retrieving subset information is acceptable, step 5
105 is unnecessary. [Effect of the Invention] As explained above, according to the present invention, when the recognition result is modified, the subset information management means operates, and the subset information in the subset information storage means is modified. can be reflected, so when the correction is reflected, the character range is limited by the fresh subse-soto information, so recognition is made in response to changes in the range of subseras 1 to 1, reducing the trouble of having to make corrections one by one. can.

[Brief explanation of drawings]

FIG. 1 is a block diagram of an embodiment of the present invention, and FIG. 2 is a flowchart for explaining the operation of an embodiment of the present invention.
FIG. 3 is a plan view of a form read by an embodiment of the present invention. 1... Photoelectric conversion unit 2... Character recognition unit 3... Subset information storage unit 4... Reading control unit 5... Operation unit 6... Display unit 21... Recognition dictionary unit 41. ... Subset information management means 410 ... Judgment section 420 ... Rewriting means

Claims (2)

[Claims] (1) A photoelectric conversion means that obtains an image signal by photoelectrically converting a form image, a recognition dictionary section that stores various parameters used during character recognition, and a subset that stores subset information that limits the range of characters involved in recognition. information storage means, and character recognition based on the image signal obtained by the photoelectric conversion means by limiting a character range using the subset information stored in the subset information storage means when necessary and referring to the parameters of the recognition dictionary section. a character recognition means for performing the above character recognition means, a modification means for modifying the recognition result of the character recognition means, and a subset information management means for performing a process of reflecting the modification by the modification means on the subset information in the subset information storage means. An optical character reading device characterized by: (2) The subset information management means includes a determining means for determining whether or not characters resulting from the correction should be added to the subset information, and rewriting the subset information in the subset information storage means in accordance with the determination result of the determining means. 2. The optical character reading device according to claim 1, further comprising rewriting means for performing rewriting.