JPH09190497A - Character recognition device - Google Patents

Abstract

(57) Abstract: To improve the recognition rate of a character recognition device. A character frame set by a character frame setting unit 5 is displayed on a tablet 1, and a preprocessing / stroke coding unit 2 linearizes a X, Y coordinate data string input from the tablet 1. After doing, code the stroke. The 1-character-frame stroke cutout unit 4 separates and outputs a plurality of stroke codes into cutout data in character frame units. The pattern matching unit 6 outputs a character code by matching the cutout data. The result character code determination unit 7 determines whether the character code is a character that may include a dakuten. The next character start stroke determination unit 8 determines whether or not there is a stroke of a voiced point of the recognition target character in the cutout data of the character to be recognized, and if there is, the recutout unit 10 integrates the voiced point. Corrected cutout data is created and the pattern matching unit 6
Give to.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a character recognition device such as an online character recognition device for identifying handwritten characters.

[0002]

2. Description of the Related Art Conventionally, in an online character recognition device, data of strokes (written portions from pen-on to pen-off) written by a person in a pen on an input device such as a tablet can be sequentially processed by a computer as XY coordinate data. Convert to a column. Then, the online character recognition device performs a cutting process on the XY coordinate data string for each of a plurality of strokes, and separates them into each character. Further, the online character recognition device performs character recognition processing on the separation result to obtain the character code of the character recognition result. The character of the character recognition result is displayed, and the character code of the character recognition result is used for various other processes. There is a pattern matching method as a general character recognition method. In this pattern matching method, feature point information is extracted from a coordinate data string of a stroke input by handwriting. The extracted information and the registered pattern information are compared by matching, and the character code is specified. here,
The registered pattern is also one in which feature points are extracted and registered in advance. In the clipping process for separating the coordinate data string for each stroke for each character, a frame for each character is set and displayed in advance on the tablet, and a method in which the writer writes in that frame, or a frame for one character in advance is used. There is a method (free format method) of setting a character writing area frame before or after character writing without setting and cutting out a coordinate data string of one character from information such as stroke intervals in the area. These methods are properly used according to the application.

[0003]

However, the conventional online character recognition device has the following problems. In the cutting process, the radical "ヽ" of the character "dog" and the semi-voiced mark """of the character" Ji "that are written outside the one-character entry frame or character area are not recognized, and" large "or ”
It is mistakenly recognized as. Also, when the punctuation marks "," are written in the same one-character frame, such as "shi,", it is erroneously recognized. That is, there is a problem that the recognition rate is lowered.

[0004]

In order to solve the above-mentioned problems, a first aspect of the present invention is based on the coordinate sequence data of the handwritten characters input into the character entry frame for each character through the coordinate input unit. A stroke encoding unit that extracts a characteristic point of a stroke that constitutes a written character and encodes the stroke according to the positional relationship of the characteristic point, and output data of the stroke encoding unit is separated for each character entry frame. A cutout unit for creating cutout data for one character by comparing the cutout data from the cutout unit with registered pattern data of characters registered in advance to extract a character code corresponding to the cutout data. A character recognition device for performing character recognition processing on the written character has the following configuration. That is, in the character recognition device of the first invention, a result character code determination unit that determines whether or not the character code corresponds to a pre-registered character code, and in the cutout data of the character next to the recognition target character , The next character start stroke determination unit that determines whether or not there is a dakuten, a semi-voiced voice, or a particular radical corresponding to the recognition target character, the determination result of the result character code determination unit, and the next character start And a re-cutout unit that corrects the separation position of the cutout unit with respect to the output data of the stroke encoding unit and creates the new cutout data based on the determination result of the stroke determination unit.

According to a second aspect of the present invention, the characteristic points of the strokes forming the written character are extracted from the coordinate sequence data of the written character input to the character entry frame for each character via the coordinate input unit, A stroke coding unit that codes a stroke according to the positional relationship of feature points, and a cutout unit that creates cutout data for one character by separating the output data of the stroke coding unit for each of the character entry frames. And a character recognition device for performing character recognition processing on the written character by comparing the cutout data from the cutout unit with registered pattern data of characters registered in advance and extracting a character code corresponding to the cutout data. And has the following configuration. That is, in the character recognition device of the second invention, the punctuation mark stroke determination unit that determines whether or not the data corresponding to the punctuation mark exists in the cutout data output from the cutout unit, and the determination of the punctuation mark stroke determination unit. As a result, when data corresponding to the punctuation mark exists in the cut-out data, the data corresponding to the punctuation mark is separated into cut-out data for one character to separate the cut-out portion from the output data of the stroke encoding unit. And a re-cutout section for correcting the above and creating new cutout data.

According to a third aspect of the present invention, the characteristic points of the strokes forming the handwritten character are extracted from the coordinate sequence data of the handwritten character input to the character entry frame for entering a plurality of characters via the coordinate input section, A stroke coding unit that codes the stroke according to the positional relationship of the feature points, and a cutout unit that creates cutout data in character units by separating the output data of the stroke coding unit into the character by character. In a character recognition device that compares the cut-out data from the cut-out portion with registered pattern data of previously registered characters to extract a character code corresponding to the cut-out data, and performs character recognition processing for the written characters. It has the following configuration. That is, in the character recognition device according to the third aspect of the present invention, the proximity determination unit that determines whether the cut-out data is the cut-out data of the character near the boundary of the character entry frame, and the determination result of the proximity determination unit is the In the case of cutout data of characters near the boundary, a result character code determination unit that determines whether or not the character code corresponding to the cutout data corresponds to a character code registered in advance, and a stroke code outside the character entry frame. Based on the judgment results of the region peripheral stroke determination unit, the result character code determination unit and the region peripheral stroke determination unit, the output of the stroke encoding unit And a re-cutout section that corrects the separation position of the cutout section with respect to the data and creates new cutout data.

According to the first aspect of the present invention, since the character recognition device is configured as described above, strokes are calculated from the coordinate sequence data of the handwritten characters input into the character entry frame for each character via the coordinate input unit. The encoding unit extracts the characteristic points of the strokes that form the written character, and encodes the strokes according to the positional relationship of the characteristic points. The cutout unit separates the output data of the stroke encoding unit for each character entry frame to create cutout data. The cutout data is compared with the registered pattern data of previously registered characters, and the character code is extracted as the recognition result. Here, the result character code determination unit determines whether or not the character code corresponds to a previously registered character code. Here, if the character code of a character that may include a dakuten, a character that may include a half-dakuten, or a character that is likely to be erroneously entered because a part of the radical is separated, the character to be recognized is determined by this determination. For example, it can be determined whether a character may include a dakuten, a character that may include a half-dakuten, or a character in which a part of the radical is separated and is likely to be miswritten. On the other hand, in the cutout data of the character next to the recognition target character, it is determined by the next character start stroke determination unit whether or not the voiced point, the semi-voiced point, or a particular radical corresponding to the recognition target character exists. It Based on the determination result of the result character code determination unit and the determination result of the next character start stroke determination unit by the re-cutout unit,
The separation position of the cutout is corrected, and the dakuten, the half-dakuten, or part of the radical erroneously entered in the next character entry frame is integrated into the cutout data of the recognition target character. The new cut-out data and the registered pattern are compared and corrected to obtain the character code.

According to the second aspect of the present invention, the punctuation mark stroke determination unit determines whether or not data corresponding to a punctuation mark exists in the cutout data output from the cutout unit. If there is data corresponding to the punctuation marks in the cutout data as a result of the judgment made by the punctuation mark stroke judgment unit, the recutting unit separates the data corresponding to the punctuation marks into cutout data for one character. That is, punctuation marks and the like that are erroneously entered together with characters in the same frame are separated and corrected as cutout data for one character. The new cutout data created by the recutout unit is compared with the registered pattern,
Correct the character code.

According to the third aspect of the invention, the stroke encoding unit extracts the characteristic points of the strokes forming the handwritten character from the coordinate sequence data of the handwritten character input in the character entry frame for entering a plurality of characters. The stroke is coded according to the positional relationship of the feature points. The output data of the stroke encoding unit is separated into each character by the cutout unit, and the character code of the character recognition result is extracted by comparing the cutout data from the cutout unit with the registered pattern data of the character registered in advance. Here, it is determined by the proximity determination means whether or not the target cutout data is the cutout data of the character near the boundary of the character entry frame, and when the determination result is the cutout data of the character near the boundary, the cutout data is processed. The result character code determination unit determines whether or not the character code that corresponds to the previously registered character code. On the other hand, whether or not there is a dakuten, a semi-dakuten, or a specific radical in the stroke code outside the character entry frame is determined by the area peripheral stroke determination unit. As a result, it is determined that the dakuten, the semi-dakuten, or a part of the radical of the character near the boundary is written outside the character entry frame. The re-cutout unit corrects the separation position of the cutout based on the determination results of the result character code determination unit and the area peripheral stroke determination unit, and creates new cutout data. The new cut-out data and the registered pattern are compared and corrected to obtain the character code. Therefore, the above problem can be solved.

[0010]

BEST MODE FOR CARRYING OUT THE INVENTION First Embodiment FIG. 1 is a functional block diagram of an online character recognition device showing a first embodiment of the present invention. This on-line character recognition device is configured by an individual circuit using an integrated circuit, or by program control of a digital signal processor (DSP), etc., and a coordinate input unit that generates position coordinates of a character input by a pen touch. Tablet 1
have. The output side of the tablet 1 is connected to the preprocessing / stroke coding section 2, and the output side of the preprocessing / stroke coding section 2 is connected to the buffer section 3. The output side of the buffer unit 3 is connected to the stroke cutout unit 4 in the one-character frame that is a cutout unit. The output side of the character frame setting unit 5 is connected to the tablet 1 and the stroke cutout unit 4 in one character frame. The pattern matching unit 6 is connected to one output side of the stroke cut-out unit 4 in one character frame, and the result character code determination unit 7 is connected to the output side of the pattern matching unit 6.

A character code such as a JIS code which is a recognition result is output from one output side of the result character code determination section 7, and the other output side of the result character code determination section 7 is The character start stroke determination unit 8 is connected. Further, the other output side of the stroke cut-out unit 4 in the one-character frame is connected to the preceding and following character stroke data storage unit 9.
, And one output side of the preceding and following character stroke data storage unit 9 is also connected to the next character start stroke determination unit 8. The output side of the next character start stroke determination unit 8 and the other output side of the preceding and following character stroke data storage unit 9 are connected to the recutting unit 10, and the output side of the recutting unit 10 is connected to the pattern matching unit 6. There is. The result character code determination unit 7, the next character start stroke determination unit 8, the preceding and following character stroke data storage unit 9, and the re-cutout unit 10 form a cutout determination unit that is a feature of this embodiment.

FIG. 2 is a flow chart showing the operation of the on-line character recognition apparatus of FIG. 1, and is shown in FIGS. 3 (a) and 3 (b).
FIG. 2 is a diagram showing a character image input from the tablet of FIG. 1 and a recognition result. 2 and 3 (a), (b)
The character recognition operation according to FIG. 1 will be described with reference to FIG. As shown in Fig. 3 (a), when there is a character entry frame for each character, the character "Ji" is not entered in one character entry frame, and a dakuten """protrudes into the adjacent character entry frame. When
By cutting out characters, "shi" and "" are separated and cut out to become "shi" and "ae". Therefore, the two characters are erroneously recognized or unrecognizable. In the first embodiment, the processing of steps S1 to S8 of FIG. 2 is performed, and the defect of the character cutout is determined by using the dakuten, the semi-voiced sound, or the radical of the character, and the cutout data of one character is corrected. To recognize. The character frame set by the character frame setting unit 5 is set on the tablet 1
Is displayed, and characters are input to the tablet 1. The character image input to the tablet 1 is X, Y
It is converted into a coordinate data string, and this X, Y coordinate data string is sequentially passed to the preprocessing / stroke coding section 2.

In step S1 of FIG. 2, the preprocessing / stroke coding unit 2 performs preprocessing on the X, Y coordinate data string given from the tablet 1, that is, removes unnecessary noise from the character image to form a straight line. Apply chemical treatment. Further, the preprocessing / stroke coding unit 2 extracts a feature point representing the feature of the handwritten character from the coordinate data string of the preprocessing result, and codes the stroke according to the positional relationship of the feature points. In other words, the preprocessing / stroke coding unit 2 separates the data string in units of strokes and performs processing for coding by stroke type. These processing results are sequentially stored in the buffer unit 3 configured by a memory device such as a RAM or a FIFO. In step S2, 1
The character frame stroke cutout unit 4 sequentially reads position (coordinate) data and stroke code from the buffer unit 3,
These are cut out in units of one character frame position (coordinates) set by the character frame setting unit 5. That is, the character frame cutout process is performed. Position data and stroke code obtained as a result of character frame cutout processing (hereinafter, these are referred to as cutout data)
Are stored in the preceding and following character stroke data storage unit 9.
The preceding and succeeding character stroke data storage unit 9 is configured to store only the cutout data for the latest two characters, and when new cutout data is input, the cutout data two characters before is deleted. In step S3 performed after step S2, the pattern matching unit 6
The feature points of the cut-out data cut out by the stroke cut-out unit 4 within one character frame are extracted, and the information of the extracted feature points is matched with the pattern information (dictionary) collected in advance to obtain the corresponding character code. Output.

In step S4, the result character code determination unit 7 determines whether the character code output by the pattern matching unit 6 is included in the character code group registered in advance. Here, as a result, the character code group used for the determination by the character code determination unit 7 includes, for example, a character that may include a dakuten, "shi", a character that may include a semi-voiced mark, "hi", "fu", and a radical. If part of the character is "dog", which is easy to be miswritten, register the character "large". When the character code output by the pattern matching unit 6 is included in the character code group, the determination result is given to the next character start stroke determination unit 8 and the process proceeds to the next step S5.
If not included, the process proceeds to step S8, and the character code output by the pattern matching unit 6 is output.
In step S5, the next character start stroke determination unit 8
Reads out the cut-out data next to the character to be recognized that is determined to be included in the character code group from the result character code determination unit 7 from the preceding and following character stroke data storage unit 9,
The stroke code near the start in the cutout data is a dakuten "", a semi-dakuten "o", or a specific radical such as "dog".
It is determined whether it corresponds to "ヽ" of. If the determination result of the next character start stroke determination unit 8 does not correspond to those dakuten, semi-dakuten, and radical, the process proceeds to step S8 to end the character recognition of the process. Proceed to step S6.

In step S6, the next character start stroke determination unit 8 determines the correspondence between the dakuten, the semi-voiced point, the particular radical and the character code from the result character code determination unit 7, which is, for example, "shi". If it is "", it is determined that they match, and the process proceeds to step S7. If "Si" and "°", it is determined that they do not match, the process proceeds to step S8, and character recognition of one character is completed. The re-cutout processing unit 10 extracts the cutout data corresponding to the character code to be recognized which is determined to be included in the character code group by the result character code determination unit 7 in step S4 and the cutout data of the next character as the preceding and succeeding characters. Stroke data storage unit 9
After reading from, the stroke data of the voiced point, the semi-voiced point, or the specific radical in the cutout data of the next character is separated, and the cutout data of the character to be recognized is combined. That is, FIG.
As in (b), the re-cutout of the character to be recognized and the next character is executed. After the re-cutout is completed, the process returns to step S3, so that the re-matching is performed.

As described above, in the first embodiment,
1 In addition to the pattern matching unit 6 that outputs the character code by the pattern matching of the cut-out data cut out by the stroke cut-out unit 4 in the character frame, the character code output by the pattern matching unit 6 is a character code in which the character code is registered in advance. As a result of determining whether the character corresponds to the character code determination unit 7 and the cutout data of the character next to the recognition target character, there is a voiced sound point, a semi-voiced sound point, or a specific radical corresponding to the recognition target character. The separation position of the stroke cutout unit 4 in one character frame is corrected based on the next character start stroke determination unit 8 that determines whether or not to perform the determination, and the determination results of the resulting character code determination unit 7 and the next character start stroke determination unit 8. Then, the re-cutout unit 10 that creates new cutout data is provided. Therefore, even if the characters "ji", "dog", etc., which should be entered in the entry frame one by one, are entered outside the frame, accurate character recognition results can be obtained and the recognition rate is improved. To do.

Second Embodiment FIG. 4 is a functional block diagram of an online character recognition device showing a second embodiment of the present invention. Like the first embodiment, this online character recognition device is also configured by an individual circuit using an integrated circuit or program control of a DSP or the like, and has a tablet 11 for inputting the position coordinates of characters with a pen touch. ing. The output side of the tablet 11 is connected to the preprocessing / stroke coding unit 12, and the output side of the preprocessing / stroke coding unit 12 is
It is connected to the buffer unit 13. The output side of the buffer unit 13 is connected to the one-character-frame stroke cutout unit 14 that is a cutout unit. The output side of the character frame setting unit 5 is connected to the tablet 11 and the stroke cut-out unit 14 within one character frame. Stroke cutout part 14 in one character frame
Is connected to the punctuation mark stroke determination unit 16, and one output side of the punctuation mark stroke determination unit 16 is
It is connected to the pattern matching unit 17. The other output side of the punctuation mark stroke determination section 16 is connected to the separated stroke storage section 18, and the output side of this separated stroke storage section 18 is connected to the recutting section 19. The output side of the re-cutout unit 19 is connected to the pattern matching unit 17, and the pattern matching unit 17 outputs a character code. The punctuation mark stroke determination unit 16, the separated stroke storage unit 18, and the recutout unit 19 constitute a cutout determination unit that is a feature of this embodiment.

FIG. 5 is a flow chart showing the operation of the on-line character recognition device of FIG. 4, and is shown in FIGS. 6 (a) and 6 (b).
FIG. 5 is a diagram showing a character image input from the tablet of FIG. 4 and a recognition result. This FIG. 5 and FIG. 6 (a), (b)
The character recognition operation according to FIG. 4 will be described with reference to FIG. As shown in FIG. 6 (a), when there is an entry frame for each character, where only the character "shi" should be entered in the letter box, a punctuation mark "," is also entered in the same box. It may cause misrecognition or unrecognition. In the second embodiment, the processing of steps S11 to S16 of FIG. 5 is performed, the deficiency of the character cutout is determined using punctuation marks, and the cutout data of one character is corrected and recognized. The tablet 11 displays the character frame set by the character frame setting unit 15, and characters are input to the tablet 11.
The character image input to the tablet 11 is converted into an X, Y coordinate data string, and this X, Y coordinate data string is sequentially passed to the preprocessing / stroke coding unit 12. In step S11 of FIG. 5, the preprocessing / stroke coding unit 12 removes noise from the character image and performs preprocessing linearization processing. Further, the preprocessing result is separated into data strings on a stroke-by-stroke basis and by stroke type. Perform coding processing. These processing results are sequentially stored in the buffer unit 13 formed of a memory device such as a RAM and a FIFO.

In step S12, the one-character-frame stroke cutout unit 14 sequentially reads the position (coordinate) data and the stroke code from the buffer unit 13, and outputs them to the one-character frame position (set by the character-frame setting unit 15). Coordinate)
Cut into units. That is, cutout data in units of one character frame is output by the same processing as that in the first embodiment. In step S13 following step S12, the punctuation mark stroke determination unit 16 determines whether or not there is a stroke corresponding to a punctuation mark in each cut-out data. If there is no stroke corresponding to a punctuation mark, the punctuation mark stroke determination unit 16 gives the cut-out data to the pattern matching unit 17 as it is and advances the process to step S15. If there is a stroke corresponding to a punctuation mark,
The punctuation mark stroke determination unit 16 inspects the position of the stroke, and if it is the position of the semi-voiced point "°", for example, the cutout data is given to the pattern matching unit 17 as it is and the process proceeds to step S15. If the position of the stroke is, for example, the position of the punctuation mark "." Or the punctuation mark ",", the punctuation mark stroke determination unit 16 separates the stroke data from the cutout data, stores it in the separated stroke storage unit 18, and the process proceeds to step S14. Proceed. In step S14, the re-cutout unit 19 outputs the separated punctuation mark strokes to the pattern matching unit 17 as cutout data of one character, as shown in FIG. 6B. In step S15, the pattern matching unit 17 matches the cutout data provided by the punctuation mark stroke determination unit 16 or the cutout data provided by the recutting unit 19, and outputs a character code in step S16.

As described above, in the second embodiment,
A punctuation mark stroke determination unit 16 that determines whether or not data corresponding to a punctuation point exists in the cutout data output by the stroke cutout unit within one character frame 14, and if data that corresponds to a punctuation point exists in the cutout data, A re-cutout section 19 for correcting the separated position of the stroke cutout section 14 in one character frame by separating it into cutout data for one character is provided, and therefore, even if punctuation marks are mixed in the character entry frame, erroneous Recognition and unrecognition do not occur and the recognition rate improves.

Third Embodiment FIG. 7 is a functional block diagram of an online character recognition device showing a third embodiment of the present invention. Like the first and second embodiments, this online character recognition device is configured by program control of an individual circuit using an integrated circuit, a DSP, etc., and a coordinate input unit for inputting the position coordinates of characters with a pen touch. It has a tablet 21 which is
The output side of the tablet 21 is connected to the preprocessing / stroke coding unit 22, and the output side of the preprocessing / stroke coding unit 22 is connected to the buffer unit 23. One output side of the buffer unit 23 is connected to the character cutout unit 24 in the character area different from those in the first and second embodiments. The character cutout unit 24 in the character area configures a cutout unit that creates cutout data in which the output data of the preprocessing / stroke coding unit 22 is separated into each character, and
Proximity determining means for determining whether or not the cutout data is cutout data of a character near the boundary of the character entry frame is configured. Tablet 21 and character cutting section 24 in the character area
An output side of the character area setting unit 25 is connected to and. On one output side of the character cutout unit 24 in the character area,
The pattern matching unit 26 is connected, and the result character code determination unit 27 is connected to the output side of the pattern matching unit 26. A character code such as a JIS code which is a recognition result is output from one output side of the result character code determination unit 27, and the area surrounding stroke determination is performed on the other output side of the result character code determination unit 27. Part 28
Is connected. The other output side of the buffer unit 23 is also connected to the area peripheral stroke determination unit 28. Further, the other output side of the character region character cutout unit 24 is connected to the region boundary vicinity cutout data storage unit 29,
Output sides of the region boundary near cutout data storage unit 29 and the region peripheral stroke determination unit 28 are connected to the recutout unit 30. The output side of the re-cutout unit 30 is connected to the pattern matching unit 26.

FIG. 8 is a flow chart showing the operation of the on-line character recognition device of FIG. 7, and is shown in FIGS. 9 (a) and 9 (b).
FIG. 9 is a diagram showing a character image input from the tablet of FIG. 7 and a recognition result. This FIG. 8 and FIG. 9 (a), (b)
The character recognition operation according to FIG. 7 will be described with reference to FIG. As shown in FIG. 9A, when there is a character entry frame for entering a plurality of characters, if the dakuten """of the character" ji "protrudes at the boundary portion of this entry frame, the recognition error occurs. In the second embodiment, the processing of steps S21 to S28 of FIG. 8 is performed, and the problem of the character cutout is determined by using the dakuten, the semi-voiced sound, or the radical of the character, and the cutout data of the character at the boundary of the entry frame. Correct and recognize. The tablet 21 is provided with a character entry frame for entering a plurality of characters set by the character area setting unit 25.
Is displayed, and characters are input to this tablet 21. The character image entered on the tablet 21 is
It is converted into an X, Y coordinate data string, and this X, Y coordinate data string is sequentially passed to the preprocessing / stroke coding section 22.

In step S21 of FIG. 8, preprocessing /
As in the first and second embodiments, the stroke encoding unit 22 removes noise from a character image and performs pre-processing linearization processing, and further, the pre-processing result is separated into data strings in stroke units and strokes. Perform processing to code by type. The results of these processes are, for example, RAM and FIF.
The data is sequentially stored in the buffer unit 23 including a memory device such as O. In step S22, the character area character cutout unit 24 sequentially reads the position (coordinate) data and the stroke code from the buffer unit 23, and corresponds to the position (coordinate) in the character entry frame set by the character area setting unit 25. Cut out each data. That is, the character region character cutout unit 24 cuts out data corresponding to each character in the character entry frame by using an algorithm generally called free format, and the obtained cutout data is sent to the pattern matching unit 26. Output. Here, the character cutout unit 24 in the character area determines whether the cutout data is for a character near the boundary of the entry frame. If the result of determination is that the cutout data is in the vicinity of the boundary, the cutout data is stored in the area boundary vicinity cutout data storage unit 29, and information indicating that the cutout data is in the vicinity of the boundary is given to the result character code determination unit 2
FIG.

In step S23, the pattern matching section 26 extracts the characteristic points of the given cut-out data, and matches the extracted characteristic point information with the pattern information (dictionary) collected in advance to correspond to the information. The character codes are output in order. In step S24, the result character code determination unit 27 determines whether the character code output from the pattern matching unit 26 at that time is a character in the vicinity of the character frame boundary, based on the information from the character region character cutout unit 24. As a result of the determination, if the character code is the character code of the character near the boundary, the processing is performed in step S.
Go to 25. If the character code is not the character near the boundary, the process proceeds to step S28, the character code is output, and the recognition of one character ends. In step S25, the result character code determination unit 27 determines whether the character code output by the pattern matching unit 26 is included in the character code group registered in advance. Here, in the character code group used for the determination by the result character code determination unit 27, for example, a character that may include a dakuten, “shi”, a character that may include a semi-voiced character, “hi,” “fu,” and one of the radicals. If the character is "dog", which is easily separated from other departments, the character such as "large" is registered. When the character code output by the pattern matching unit 26 is included in the character code group, the determination result is given to the area stroke determination unit 28, and the process proceeds to the next step S26. If not included, the process proceeds to step S28, the character code is output, and the recognition of one character ends.

In step S26, the area peripheral stroke determination section 28 reads the stroke code next to the character determined in step S25 from the buffer section 23 which also stores data outside the character frame. The area peripheral stroke determination unit 28 determines whether or not the stroke code read from the buffer unit 23 corresponds to a dakuten """, a semi-dakuten" ° ", or" ヽ "of a specific radical, such as" dog ". To do. If the stroke code does not correspond to those dakuten, semi-dakuten, and radical as a result of the determination, the area surrounding stroke determination unit 28 advances the processing to step S28, terminates the character recognition of the processing, and if they correspond, Takes the correspondence between the stroke code and the character to be judged as in the first embodiment, and advances the processing to step S27. Step S
In 27, the re-cutout processing unit 30 combines the stroke code from the region peripheral stroke determination unit 28 and the cutout data of the determination target character from the region boundary near cutout data storage unit 24, and as shown in FIG. 9B. The new cut-out data in which the dakuten, the semi-dakuten, or a specific radical is captured is formed and output to the pattern matching unit 26, and the process is step S.
Return to 23. Re-matching is performed by returning the process to step S3.

As described above, in the third embodiment,
Characters in a character area for generating cutout data in which the output data of the preprocessing / stroke coding unit 22 is separated character by character and determining whether the cutout data is cutout data of a character near the boundary of the character entry frame Cutout part 24
In the case of the cut-out data of the character near the boundary, the result character code judgment unit 27 for judging whether or not the character code corresponding to the cut-out data corresponds to the character code registered in advance, and the stroke code outside the character entry frame. Based on the judgment results of the region peripheral stroke determination unit 28, the result character code determination unit 27, and the region peripheral stroke determination unit 28, the character region character segmentation is performed based on the determination results of the region surrounding stroke determination unit 28 and the result character code determination unit 27. A re-cutout unit 30 that corrects the separation position of the unit 24 and creates new cutout data is provided. Therefore, even if there is a dakuten, a half-dakuten, or a radical that has been entered outside the character entry frame for entering a plurality of characters, it is possible to perform character recognition incorporating it and the recognition rate is improved.

The present invention is not limited to the above embodiment, and various modifications can be made. For example, there are the following modifications. (1) In the first to third embodiments, the tablets 1, 1
Although an example in which the present invention is applied to an online character recognition device using 1 and 21, the present invention is also applied to a device that optically reads a character image and cuts out a character to perform character recognition, such as an OCR device. Is possible. (2) In the first to third embodiments, the mode in which the character image is input after the character entry frame is set in advance has been described. However, the character area frame targeted for clipping or character recognition is input as the character image. Even if you set it later,
The same effect as the third embodiment can be obtained.

[0028]

As described in detail above, according to the first aspect of the present invention, a stroke encoding unit for encoding strokes from coordinate sequence data of handwritten characters input in a character entry frame for each character. And a cutout unit for creating cutout data, and the result character code is applied to a character recognition device that performs character recognition processing on the written character by extracting the character code by comparing the cutout data with the registered pattern data of the character registered in advance. Since the determination unit, the next character start stroke determination unit, and the re-cutout unit are provided, the cutout data is corrected, and the "ji" of the character to be entered in the entry frame one by one,
Even when "dog" or the like is entered outside the frame, an accurate character recognition result can be obtained and the recognition rate is improved.
According to the second invention, a stroke coding unit that codes a stroke from the coordinate sequence data of the handwritten characters that are respectively input to the character entry frames for each character, and a cutout unit that creates cutout data are provided, A character recognition device that performs character recognition processing for handwritten characters by extracting character codes by comparing cutout data with registered pattern data of previously registered characters is provided with a punctuation mark stroke determination unit and recutout unit. Even if punctuation marks are mixed in the frame, they are separated, no erroneous recognition or unrecognizable occurs, and the recognition rate is improved. According to the third aspect of the present invention, a stroke encoding unit that encodes a stroke from the coordinate sequence data of the written characters input in the character entry frame for entering a plurality of characters, and a cutout unit that creates cutout data in character units. In the character recognition device for extracting the character code by comparing the cutout data with the registered pattern data of the character registered in advance and performing the character recognition process for the written character, the proximity determination means, the result character code determination unit, and the area peripheral stroke Since the judgment unit and the re-cutout unit are provided, even if there are dakuten, semi-voiced marks or radicals that have been entered outside the character entry frame for entering multiple characters, cutout data that captures them can be obtained. , Recognition rate improves

[Brief description of the drawings]

FIG. 1 is a functional block diagram of an online character recognition device showing a first embodiment of the present invention.

FIG. 2 is a flowchart showing an operation of the online character recognition device of FIG.

FIG. 3 is a diagram showing a character image input from the tablet of FIG. 1 and a recognition result.

FIG. 4 is a functional block diagram of an online character recognition device showing a second embodiment of the present invention.

5 is a flowchart showing an operation of the online character recognition device of FIG.

6 is a diagram showing a character image input from the tablet of FIG. 4 and a recognition result.

FIG. 7 is a functional block diagram of an online character recognition device showing a third embodiment of the present invention.

FIG. 8 is a flowchart showing the operation of the online character recognition device of FIG.

9 is a diagram showing a character image input from the tablet of FIG. 7 and a recognition result.

[Explanation of symbols]

1,11,21 Tablets 2,12,22 Pre-processing / stroke coding unit 3,13,23 Buffer unit 4,14 1 Character stroke cutout unit 5,15 Character frame setting unit 6,17,26 Pattern matching unit 7, 27 Result character determination unit 8 Next character start stroke determination unit 9 Previous character stroke data storage unit 10, 19, 30 Re-cutout unit 16 Punctuation stroke determination unit 18 Separation stroke storage unit 24 Character cutout unit in character area 25 Character area setting unit 28 Area Perimeter Stroke Judgment Section 29 Area Border Neighborhood Cutout Data Storage Section

─────────────────────────────────────────────────── ─── Continuation of front page (51) Int.Cl. ⁶ Identification code Internal reference number FI Technical display location G06K 9/62 640 9061-5H G06K 9/62 640A 9061-5H G G06F 17/22 G06F 15/20 506A

Claims (3)

[Claims] 1. A feature point of a stroke forming the handwritten character is extracted from the coordinate sequence data of the handwritten character input into the character entry frame for each character via the coordinate input unit, and the position of the feature point is extracted. A stroke coding unit that codes the stroke according to the relationship; and a cutout unit that creates cutout data for one character by separating the output data of the stroke coding unit for each of the character entry frames, In a character recognition device that compares the cut-out data from the cut-out portion with registered pattern data of previously registered characters to extract a character code corresponding to the cut-out data, and performs character recognition processing for the written characters, the character As a result of determining whether the code corresponds to a pre-registered character code, a character code determination unit, and in the cutout data of the character next to the recognition target character , A next character start stroke determination unit that determines whether or not there is a dakuten, a semi-voiced voice, or a particular radical of the character to be recognized, the determination result of the result character code determination unit, and the next character start stroke And a re-cutout section that corrects the separation position of the cutout section with respect to the output data of the stroke encoding section to create new cutout data based on the determination result of the determination section. Recognition device. 2. A feature point of a stroke that constitutes the written character is extracted from the coordinate sequence data of the written character that is input in the character entry frame for each character via the coordinate input unit, and the position of the feature point is extracted. A stroke coding unit that codes the stroke according to the relationship; and a cutout unit that creates cutout data for one character by separating the output data of the stroke coding unit for each of the character entry frames, In a character recognition device that compares the cutout data from the cutout unit with registered pattern data of preregistered characters to extract a character code corresponding to the cutout data and performs a character recognition process for the written character, the cutout A punctuation mark stroke determination unit that determines whether or not data corresponding to a punctuation mark exists in the cutout data output from the unit; If data corresponding to the punctuation mark is present in the cut-out data as a result of the judgment by the judgment unit, the data corresponding to the punctuation mark is separated into the cut-out data for one character, and the data corresponding to the output data of the stroke encoding unit is separated. And a re-cutout unit that corrects the separation position of the cutout unit and creates new cutout data. 3. A characteristic point of a stroke forming the written character is extracted from the coordinate sequence data of the written character input to the character entry frame for writing a plurality of characters via the coordinate input unit, and the position of the characteristic point is extracted. A stroke coding unit that codes the stroke according to a relationship; and a cutout unit that creates cutout data in character units that separates the output data of the stroke coding unit into the one character at a time. In the character recognition device for comparing the cutout data of the above with the registered pattern data of the previously registered characters to extract the character code corresponding to the cutout data and performing the character recognition process for the written character, the cutout data is the character Proximity determination means for determining whether or not the data is the cut-out data of the characters near the boundary of the entry frame, and the determination result of the proximity determination means is In the case of output data, the result character code determination unit that determines whether or not the character code corresponding to the cut-out data corresponds to a character code registered in advance, and a stroke code outside the character entry frame A region peripheral stroke determination unit that determines whether or not there is a dakuten or a particular radical, and based on the determination results of the result character code determination unit and the region peripheral stroke determination unit, the cutout for the output data of the stroke encoding unit. A character recognizing device, comprising: a re-cutout unit that corrects the separation position of the copy and creates the new cutout data.