JP2002109474A - Character recognition device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a character recognition device which processes recognition efficiently and has high accuracy of recognition. SOLUTION: An online recognition (S4) and a rough offline recognition (S6) are processed on a handwriting character which is inputted on a tablet for handwriting input character, and the result of the recognition is integrated (S8). When the resemblance of the integrated result is over the threshold (S10: YES), it is displayed on the display. When the result is under the threshold (S10: NO), offline recognition is processed in detail again (S12), and the result of that is displayed on the display.

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 that recognizes input handwritten characters and converts them into character codes.

[0002]

2. Description of the Related Art Conventionally, methods for recognizing handwritten characters include:
One line (hereinafter, referred to as “stroke”), which is a constituent line of an input character, is used as a recognition unit, and online recognition is performed using a time series of the input order of strokes. Off-line recognition, which performs recognition using shapes, is known. Online recognition has a feature that the recognition process is performed when a stroke is input, so the recognition process of handwritten characters is fast.However, the stroke order and strokes such as continuation characters, overwriting, etc. It has the disadvantage of being brittle. In addition, the offline recognition has a feature that since the recognition is performed using the shape of the entire input character, there is little erroneous recognition due to a stroke order error or overwriting, but the overall balance and stroke of the input character are small. Is disadvantageous in that it is brittle when the position is shifted.

However, as described above, Japanese Patent Laid-Open Publication No. Hei 9-288717 discloses an online recognition and an offline recognition as means for improving the disadvantages of the online recognition and the offline recognition because their features are mutually complementary as described above. Character recognition by a hybrid method combining online recognition and offline recognition has been proposed. The character recognition by the hybrid method has an advantage that the accuracy of the recognition is higher than that of performing any one of the online recognition and the offline recognition because the disadvantages of the online recognition and the offline recognition are complemented. .

[0004]

However, depending on the input contents of the handwritten characters, there is a case where a sufficiently reliable result can be obtained only by the online recognition or only the offline recognition. Even in such a case, the character recognition by the hybrid method is required. Doing so results in excessive recognition processing, and there is a problem that the character recognition processing takes time and is inefficient.

SUMMARY OF THE INVENTION The present invention has been made to solve the above-described problems, and has as its object to provide a character recognition device which has high recognition processing efficiency and high recognition accuracy.

[0006]

According to a first aspect of the present invention, there is provided a character recognition apparatus for recognizing an input handwritten character and converting the input handwritten character into a character code. Handwritten character input means for inputting a character, first character recognition means for recognizing the handwritten character input from the handwritten character input means and converting it to a character code, and handwritten character input from the handwritten character input means Recognize and convert to a character code, and a second character recognition means having higher recognition accuracy of handwritten characters than the first character recognition means, and a recognition candidate recognized by the first character recognition means A similarity determination unit configured to determine a similarity to a basic character. The similarity determination unit determines a similarity of the recognition candidate recognized by the first character recognition unit. Is greater than or equal to a predetermined value, the recognition candidate is output as a recognition result. If the similarity is less than the predetermined value, character recognition is performed again by the second character recognition means, and the recognition candidate is It is configured to output as a recognition result.

In the character recognition device having this configuration, when the similarity is judged by the similarity judgment means on the recognition candidates recognized by the first character recognition means, and the similarity is equal to or more than a predetermined value, the recognition is performed. A candidate is output as a recognition result, and if the similarity is less than a predetermined value, character recognition is performed again by the second character recognition unit having higher recognition accuracy of the handwritten character than the first character recognition unit, and the recognition candidate Is output as the recognition result.

According to a second aspect of the present invention, in addition to the configuration of the character recognition apparatus of the first aspect,
The second character recognizing unit, when the similarity of the recognition candidate recognized by the first character recognizing unit is less than a predetermined value, the recognition candidate output from the first character recognizing unit. Character recognition is performed again for those having a high similarity to a predetermined rank, and the recognition candidate is output as a recognition result.

[0009] In the character recognition device having this configuration, in addition to the operation of the character recognition device according to the first aspect, the second character recognition means includes a similarity measure of the recognition candidates recognized by the first character recognition means. Is less than a predetermined value, the character recognition is again performed for recognition candidates output from the first character recognition unit for those having a high similarity to a predetermined order,
The recognition candidate can be output as a recognition result.

According to a third aspect of the present invention, in addition to the configuration of the character recognition device according to the first or second aspect, the second character recognition unit includes the first character recognition unit. As compared with the above, the configuration is characterized in that handwritten characters are recognized using a large amount of character feature data.

[0011] In the character recognition device having this configuration, in addition to the operation of the character recognition device according to claim 1 or 2, the second character recognition means is more effective than the first character recognition means. Handwritten characters can be recognized using a large amount of feature data.

According to a fourth aspect of the present invention, in addition to the configuration of the character recognition device according to any one of the first to third aspects, the first character recognition unit includes the second character recognition unit. The configuration is characterized in that the processing speed of character recognition is faster than the character recognition means.

In the character recognition device having this configuration, in addition to the operation of the character recognition device according to any one of claims 1 to 3, the first character recognition means has a higher speed than the second character recognition means. Can perform character recognition.

According to a fifth aspect of the present invention, in addition to the configuration of the character recognition device according to any one of the first to fourth aspects, the first character recognition unit includes the handwritten character input device. It is characterized by performing online recognition for recognizing handwritten characters using time-series information of constituent lines of handwritten characters input from the means, the second character recognition means,
Off-line recognition for recognizing a handwritten character using the entire shape of the handwritten character input from the handwritten character input unit is performed.

In the character recognition device having this configuration, in addition to the operation of the character recognition device according to any one of claims 1 to 4, the first character recognition means includes a handwritten character input from the handwritten character input means. Performs online recognition for recognizing handwritten characters using time-series information of character constituent lines, and the second character recognition unit uses the entire shape of the handwritten characters input from the handwritten character input unit. Offline recognition for recognizing handwritten characters can be performed.

According to a sixth aspect of the present invention, in addition to the configuration of the character recognition device according to any one of the first to fifth aspects, the first character recognition means includes the handwritten character input. Means for recognizing handwritten characters using time-series information of constituent lines of the handwritten characters input from the means, and handwriting using the entire shape of the handwritten characters input from the handwritten character input means. It is characterized by generating a new recognition candidate based on the recognition result of the offline recognition for character recognition, wherein the second character recognition means, compared to the offline recognition performed by the first character recognition means, Offline recognition is performed using a large amount of character feature data.

In the character recognition device having this configuration, in addition to the operation of the character recognition device according to any one of claims 1 to 5, the first character recognition means includes a handwritten character input from the handwritten character input means. Recognition results by online recognition that recognizes handwritten characters using time-series information of the constituent lines of
A new recognition candidate can be generated based on the recognition result of the offline recognition for recognizing the handwritten character using the entire shape of the handwritten character input from the handwritten character input means, and the second character recognition means includes: Compared to off-line recognition performed by one character recognition unit, off-line recognition can be performed using a large amount of character feature data.

According to a seventh aspect of the present invention, there is provided a character recognition apparatus for recognizing an input handwritten character and converting the input handwritten character into a character code. First character recognition means for recognizing a handwritten character input from the handwritten character input means and converting it into a character code, wherein the first character recognition means comprises a handwritten character input from the handwritten character input means. Online recognition means for recognizing handwritten characters using the time-series information of the constituent lines, and offline recognition means for recognizing handwritten characters using the entire shape of the handwritten characters input from the handwritten character input means. Specifying whether to perform character recognition of the handwritten character input from the handwritten character input unit by either the online recognition unit or the offline recognition unit. And a recognition unit designating means, the recognition means which is specified by said recognition means designating means has a configuration which is characterized in that the recognition of handwritten characters input from said handwriting input means.

In the character recognition device having the above configuration, the first character recognition means includes an online recognition means for recognizing handwritten characters using time-series information of constituent lines of the handwritten characters input from the handwritten character input means; Offline recognition means for recognizing handwritten characters using the entire shape of the handwritten characters input from the handwritten character input means; and handwriting characters input from the handwritten character input means by either the online recognition means or the offline recognition means. Since there is provided recognition means designating means for designating whether or not to perform character recognition, the recognition means designated by the recognition means designating means can recognize handwritten characters input from the handwritten character input means.

According to the character recognition device of the invention according to claim 8, in addition to the configuration of the character recognition device according to claim 7,
The first character recognition means is an online recognition means for recognizing a handwritten character using time-series information of constituent lines of the handwritten character input from the handwritten character input means, and an input from the handwritten character input means. At least an off-line recognition unit for recognizing a handwritten character using the entire shape of the handwritten character, and a recognition candidate generation unit for generating a new recognition candidate based on a recognition result of the on-line recognition unit and the off-line recognition unit. Recognizing means specifying which of the two means performs character recognition of the handwritten character input from the handwritten character input means, and a recognizing means specified by the recognizing means specifying means. , And recognizes handwritten characters input from the handwritten character input means.

[0021] In the character recognition device having this configuration, in addition to the operation of the character recognition device according to claim 7, the first character recognition means is configured to determine a constituent line of a handwritten character input from the handwritten character input means. Online recognition means for recognizing handwritten characters using time-series information; offline recognition means for recognizing handwritten characters using the entire shape of the handwritten characters input from the handwritten character input means; Means for generating a new recognition candidate based on the recognition result of the off-line recognition means, and at least two of the recognition candidate generating means for generating a new recognition candidate. And recognition means designating means for designating whether or not to perform the handwriting by the recognition means designated by the recognition means designating means. It can be recognized handwriting input from the character input means.

According to a ninth aspect of the present invention, in addition to the configuration of the character recognition apparatus according to the seventh or eighth aspect, the first character recognition means includes an input means for inputting from the handwritten character input means. Online recognition means for recognizing handwritten characters by using the time-series information of the constituent lines of the handwritten characters, and handwritten character recognition using the entire shape of the handwritten characters input from the handwritten character input means Offline recognition means, recognition candidate generation means for generating new recognition candidates based on the recognition results of the online recognition means and the offline recognition means, the online recognition means or the offline recognition means or the online recognition means and the Designating which of the offline recognition means performs character recognition of the handwritten character input from the handwritten character input means A identification means specifying means, the recognition means which is specified by said recognition means designating means has a configuration which is characterized in that the recognition of handwritten characters input from said handwriting input means.

[0023] In the character recognition device having this configuration, in addition to the operation of the character recognition device according to claim 7 or 8, the first character recognition means includes a configuration of a handwritten character input from the handwritten character input means. Online recognition means for recognizing handwritten characters using time-series information of lines, offline recognition means for recognizing handwritten characters using the entire shape of the handwritten characters input from the handwritten character input means, A recognition candidate generating unit that generates a new recognition candidate based on the recognition result of the online recognition unit and the offline recognition unit; and any of the online recognition unit or the offline recognition unit or the online recognition unit and the offline recognition unit. Recognition means designating means for designating whether to perform character recognition of handwritten characters input from the handwritten character input means Is provided with the, recognition means which is specified by said recognition means specifying means, it is possible to perform the recognition of handwritten characters input from said handwriting input means.

According to a tenth aspect of the present invention, in addition to the configuration of the character recognition apparatus according to any one of the first to ninth aspects, the character recognition apparatus recognizes a handwritten character input from the handwritten character input means. Display means for displaying the recognized recognition candidates.

According to the character recognition device having this configuration, in addition to the operation of the character recognition device according to any one of claims 1 to 9, a recognition unit that recognizes a handwritten character input from the handwritten character input unit by the display unit. Candidates can be displayed.

[0026]

DESCRIPTION OF THE PREFERRED EMBODIMENTS One embodiment of the present invention will be described below with reference to the drawings.

FIG. 1 is a block diagram showing a schematic configuration of the character recognition device 1, FIG. 2 is a plan view of a handwritten character input tablet 4, and FIG. 3 is a plan view of a display device 6.

As shown in FIG. 1, a character recognition device 1 according to a first embodiment of the present invention comprises a RAM 8 for temporarily storing data in a CPU 2 for controlling the character recognition device 1,
A ROM 7 storing a program to be executed in U2,
An I / O interface 3 that mediates input / output of data to / from the CPU 2 is connected. The I / O interface 3 is connected to a handwritten character input tablet 4 for inputting characters handwritten by the pen 5 and a display device 6 including a liquid crystal display device.

Next, the storage area of the ROM 7 will be described. As shown in FIG. 1, the ROM 7 has a character extraction program storage area 7a, an online recognition program storage area 7b, and an online recognition dictionary storage area 7c.
, An off-line recognition program storage area 7d, off-line recognition dictionary storage areas 7e and 7f, a recognition result integrated program storage area 7g, and a primary recognition result availability determination threshold storage area 7h.

The character extraction program storage area 7a
Stores a character extraction program for causing the CPU 2 to execute a process of identifying one character from data input from the handwritten character input tablet 4 and extracting the data as one character data. In the online recognition program storage area 7b, one line (stroke), which is a constituent line of a character input to the handwritten character input tablet 4, is used as a recognition unit, and recognition is performed using a time series of the stroke input order. An online recognition program that causes the CPU 2 to execute the online recognition process to be performed is stored. Further, the online recognition dictionary storage area 7c stores an online recognition dictionary referred to in the online recognition processing.
In this online recognition dictionary, all strokes of the character are stored for each character.

In the off-line recognition program storage area 7d, there is stored an on-line recognition program for causing the CPU 2 to execute an off-line recognition process for recognizing the whole character of one character input to the handwritten character input tablet 4. ing. Further, in the offline recognition dictionary storage area 7e, an offline recognition dictionary in which data of the shape of one character to be referred to in the offline recognition processing is composed of 16-dimensional (16) feature data is stored. In addition, the offline recognition dictionary storage area 7f stores an offline recognition dictionary in which data of the shape of one character referred to in the offline recognition processing is composed of 64-dimensional (64) feature data. The recognition result integration program storage area 7g stores a recognition result integration program that causes the CPU 2 to execute processing for integrating the results of online recognition and offline recognition. In the primary recognition result availability determination threshold storage area 7h, a process (S10) for determining a primary recognition result, which will be described later, is performed by the CPU 2.
Is stored as a threshold value for the determination.

Next, the storage area of the RAM 8 will be described. As shown in FIG. 1, the RAM 8 stores an input stroke data storage area 8a for storing strokes input to the handwritten character input tablet 4, and a result of online recognition of characters input to the handwritten character input tablet 4. An online recognition result storage area 8b to be stored;
Offline recognition result storage area 8c for storing the result of off-line recognition of characters input to handwritten character input tablet 4, and the result of recognizing characters input to handwritten character input tablet 4 using online recognition and offline recognition , And a working area 8e for storing data in various processes performed by the CPU 2.

Next, the structure of the handwritten character input tablet 4 will be described with reference to FIG. As shown in FIG. 2, the handwritten character input tablet 4 is formed in a horizontally long rectangle, and a first input area 4 is provided on the surface of the handwritten character input tablet 4 as an input area for inputting handwritten characters.
a, a second input area 4b, and a third input area 4c are provided. The first input area 4a, the second input area 4b, and the third input area 4c are each slightly longer in length than in width. It is formed in a rectangle, and three are arranged adjacently. The first input area 4a and the second input area 4
(b) When characters are written in any of the input areas of the third input area 4c with the pen 5, the input stroke data is sent to the CPU 2 via the I / O interface 3.

Further, recognition mode switches 4d, 4e and 4f for designating a primary recognition mode are provided at the upper left of the handwriting character input tablet 4. When the recognition mode switch 4d is pressed with the pen 5, the primary recognition of the handwritten characters input to the handwritten character input tablet 4 includes online recognition described later (S4 in FIG. 5) and coarse offline recognition described later (S6 in FIG. 5). The recognition mode is stored in the RAM working area 8e so as to perform both. When the recognition mode switch 4e is pressed with the pen 5, the recognition mode is set to RAM working so that only online recognition (S4 in FIG. 5) described later is performed as primary recognition of handwritten characters input to the handwritten character input tablet 4. It is stored in the area 8e. When the recognition mode switch 4f is pressed with the pen 5, the recognition mode is set to RAM so that only coarse offline recognition (S6 in FIG. 5) described later is performed as primary recognition of handwritten characters input to the handwritten character input tablet 4. It is stored in the working area 8e. Note that the handwritten character input tablet 4 may incorporate a liquid crystal and use a liquid crystal tablet on which an input stroke is displayed.

The inside of the handwritten character input tablet 4 is provided with two conductive films having a resistance of several hundred Ω in parallel with the surface of the handwritten character input tablet 4 with a slight gap therebetween. Tablet for input 4
When the surface of is pressed by the tip of the pen 5, the two conductive films come into contact with each other and the X coordinate and the Y coordinate are read.

Next, the structure of the display device 6 will be described with reference to FIG. The display device 6 is formed in a horizontally long rectangle, and a display unit 6a formed of a liquid crystal display device (LCD) is provided on a surface of the display device 6. The display unit 6a includes:
Recognized characters are displayed in the order of input.

The operation of the character recognition device 1 configured as described above will be described with reference to FIGS. FIG.
5 is a flowchart of a character recognition process performed by the character recognition device 1.

Hereinafter, a case where the user inputs a character "sentence" using the pen 5 will be described as an example. Here, the recognition mode switch 4d of the handwritten character input tablet 4 is depressed with the pen 5, and the primary recognition of the handwritten characters input to the handwritten character input tablet 4 includes online recognition described later and coarse offline recognition described later. It is assumed that the recognition mode is set to perform both. First, when the user uses the pen 5 to enter a character of “sentence” in any of the input areas of the first input area 4a to the third input area 4c of the tablet 4 for inputting handwritten characters (S1), The stroke of each stroke in which the character "" is handwritten is stored in the input stroke data storage area 8a of the RAM 8 together with the time-series information (S2).

At this time, it is determined according to the character extraction program stored in the character extraction program storage area 7a of the ROM 7 whether or not the input of one character has been completed (S3). The determination as to whether or not the input of one character is completed may be made, for example, when the time until the input of the next stroke is opened for a fixed time (for example, “1 second” or “2 seconds”) or when the input is performed in the first input area 4a. What was done is the second input area 4b
It is determined that the input of one character has been completed, for example, when input has been started in the second input area 4b or when input has been started in the third input area 4c (S3: YE).
S). If the input of one character has not been completed (S3:
NO), and return to the process of S1.

When the input of one character is completed (S3:
Then, online recognition is performed (S4). Specifically, based on the stroke information stored in the input stroke data storage area 8a, online recognition is performed with reference to the online recognition dictionary stored in the online recognition dictionary storage area 7c of the RAM 7 (S4).
Here, the CPU 2 executes the online recognition program stored in the online recognition program storage area 7b of the ROM 7, thereby storing the time-series information of strokes stored in the online recognition dictionary and the input stroke data storage area 8a. The character is identified by comparing the time series information of the stroke
4).

For example, as a result of performing online recognition in S4 for the input of a “sentence” of handwritten characters, “six” as a first candidate, “sentence” as a second candidate, and a third If the candidate "father" is generated as a candidate, the recognition result is stored in the online recognition result storage area 8b of the RAM 8 (S5).

Next, rough offline recognition is performed (S
6). Specifically, the input stroke data storage area 8
a based on the stroke information stored in the
The coarse offline recognition is performed with reference to the 16-dimensional offline recognition dictionary stored in the offline recognition dictionary storage area 7e of the first step (S6). Here, the CPU 2 is the RAM 7
By executing the off-line recognition program stored in the off-line recognition program storage area 7d of
The character is identified by comparing the information on the shape of the character stored in the six-dimensional off-line recognition dictionary with the information on the shape of the character composed of the strokes stored in the input stroke data storage area 8a (S6). .

Here, what is referred to as "coarse offline recognition" is that the characteristic data of the character shape used for offline recognition is composed of 16-dimensional (16) characteristic data, This is because the recognition processing speed is faster than the detailed off-line recognition using (64) feature data, but the recognition accuracy is low. As a result of the coarse offline recognition (S6), for example, in response to the input of the "sentence" of handwritten characters, "interchange" as the first candidate, "sentence" as the second candidate, and "sentence" as the third candidate If the candidate "no" is generated, the recognition result is stored in the off-line recognition result storage area 8c of the RAM 8 (S7).

Next, the online recognition result storage area 8
The candidate of the online recognition result stored in b and the candidate of the offline recognition result stored in the offline recognition result storage area 8c are integrated (S8). Here, first, a recognition score is given to each of the online recognition result and the coarse offline recognition result. Specifically, a value obtained by normalizing the similarity (distance) between the recognized candidate character and the data used for recognition with a value of “0 to 100” is used. For example,
In online recognition, the closer the recognition score is to "100", the higher the matching rate between the time-series information of the stroke stored in the input stroke data storage area 8a and the time-series information of the stroke of the recognition candidate character increases. The closer the score is to “0”, the more the input stroke data storage area 8
The coincidence rate between the time-series information of the stroke stored in “a” and the time-series information of the stroke of the character of the recognition candidate becomes low.
In the offline recognition, the closer the recognition score is to “100”, the higher the matching rate between the characteristic information of the shape of the character composed of the strokes stored in the input stroke data storage area 8a and the characteristic information of the character shape of the recognition candidate. Is higher, and the closer the recognition score is to “0”, the lower the matching rate between the characteristic information of the shape of the character composed of the strokes stored in the input stroke data storage area 8a and the characteristic information of the character shape of the recognition candidate is lower. Become.

In this way, the candidates for the online recognition result with the recognition score are, for example, “six (90)”, “sentence (85)”, and “father (60)”. For example, "intersection (85)", "sentence (8
0) "and" no (40) ". Then, when both recognition scores are added for each character and arranged in descending order of the total recognition score, “sentence (165)”, “six (90)”, “interchange (85)”, “father (60)”, "No (40)"
The result of this integration is “sentence”, “six”, “interchange”, “father”,
In the order of "no", the information is stored in the hybrid recognition result storage area 8d together with the respective total recognition scores (S9).

Next, it is determined whether or not the similarity is equal to or larger than a threshold (S10). In the determination process of S10, the total recognition score (similarity) of the characters of the integration result candidates integrated by the recognition result integration process of S8 and stored in the hybrid recognition result storage area 8d is determined by the primary recognition result availability determination threshold of the ROM 7. It is determined whether the value is equal to or greater than a predetermined threshold value stored in the storage area 7h. For example, the predetermined threshold stored in the primary recognition result availability determination threshold storage area 7h is “160”.
If, the candidate of the integration result of the total recognition score (similarity) equal to or larger than the threshold is stored in the working area 8e functioning as the display buffer of the RAM 8 (S11: YES) (S11). Here, the total recognition score (similarity) of “sentence”, which is the first place of the candidate character of the integration result stored in the hybrid recognition result storage area 8d, is “165” and is equal to or more than the threshold “160”. (S10: YE
S), the character of "sentence" is stored in the working area 8e (S11), and the display unit 6a of the display device 6 as shown in FIG.
Is displayed (S15). Then, the handwritten character recognition processing ends.

In the determination processing of S10, the total recognition score of the characters of the integration result candidates integrated by the recognition result integration processing of S8 and stored in the hybrid recognition result storage area 8d is stored in the primary recognition result availability determination threshold storage area of the ROM 7. 7
The predetermined threshold value stored in h (for example, “160”)
If it is less than 1, it is considered that the recognition accuracy is poor, that is, the reliability is low. Therefore, detailed offline recognition using 64-dimensional (64) feature data is performed (S12).

More specifically, of the characters of the integration result candidates stored in the hybrid recognition result storage area 8d, those from the highest total recognition score (similarity) to the predetermined order are input stroke data storage area 8a. The detailed off-line recognition is performed with reference to the 64-dimensional off-line recognition dictionary stored in the off-line recognition dictionary storage area 7f of the ROM 7 based on the stroke information stored in (1) (S12).

In the detailed offline recognition process (S12), the CPU 2 executes the offline recognition program stored in the offline recognition program storage area 7d of the RAM 7 to obtain the shape of the character stored in the 64-dimensional offline recognition dictionary. The character is identified by comparing this information with the information on the shape of the character composed of the stroke stored in the input stroke data storage area 8a (S12). In this detailed offline recognition, since offline recognition is performed using 64-dimensional (64) feature data, offline recognition can be performed using feature data four times larger than the coarse offline recognition described above. Is longer than coarse offline recognition, but has the characteristic that recognition accuracy is high.

When the detailed off-line recognition processing (S12) is completed, the recognition result is stored in the off-line recognition result storage area 8c (S13), and the candidate character of the recognition result stored in the off-line recognition result storage area 8c is stored in the RAM 8 It is stored in the working area 8e functioning as a display buffer (S14), and is displayed on the display unit 6a of the display device 6 as shown in FIG. 3 (S15). Then, the handwritten character recognition processing ends.

As described above, in the character recognition device 1 according to the first embodiment of the present invention, first, online recognition and coarse offline recognition are performed as primary recognition. Processing speed can be increased. Also, if the similarity of the result of integrating the online recognition and the coarse offline recognition is equal to or greater than a predetermined threshold, it is determined that the recognition accuracy is high, and the integrated recognition candidates are displayed as they are, so that the speed of the recognition processing is reduced. The accuracy of the recognition result can be kept constant while increasing. In addition, if the similarity of the result of integrating the online recognition and the coarse offline recognition is less than a predetermined threshold, the recognition accuracy is determined to be low, and the detailed offline recognition process is performed again. There is no end.

Next, a second embodiment of the present invention will be described with reference to the flowchart shown in FIG. FIG.
9 is a flowchart of a character recognition process of the character recognition device according to the second embodiment. The second embodiment differs from the first embodiment only in the character recognition process shown in FIG. 5 and is completely the same in other mechanical and electrical configurations. Description will be made with reference to the reference numerals of the respective components.

In the second embodiment, the recognition mode switch 4e of the handwritten character input tablet 4 is depressed with the pen 5, and the primary recognition of the handwritten characters input to the handwritten character input tablet 4 is only online recognition. It is assumed that the recognition mode is set so as to perform First, when the user uses the pen 5 to input a "sentence" character into any of the first to third input areas 4a to 4c of the handwritten character input tablet 4 (S21). The stroke of each one-stroke handwritten with the word "sentence" is stored in the input stroke data storage area 8a of the RAM 8 together with the time-series information (S22). At this time, it is determined whether or not the input of one character has been completed according to the character extraction program stored in the character extraction program storage area of the ROM 7 (S23). The determination as to whether or not the input of one character is completed may be made, for example, when the time until the input of the next stroke is opened for a fixed time (for example, “1 second” or “2 seconds”) or when the input is performed in the first input area 4a. It is assumed that the input of one character is completed, for example, when the input has been started in the second input area 4b, or the input in the second input area 4b has started input in the third input area 4c. A determination is made (S23: YES). If the input of one character has not been completed (S23: NO), the process returns to S21.

When the input of one character is completed (S2
3: YES), and then perform online recognition (S2)
4). Specifically, the input stroke data storage area 8
a based on the stroke information stored in the RAM 7
The online recognition is performed with reference to the online recognition dictionary stored in the online recognition dictionary storage area 7c (S24). Here, the CPU 2 executes the online recognition program stored in the online recognition program storage area 7b of the ROM 7, thereby storing the time-series information of strokes stored in the online recognition dictionary and the input stroke data storage area 8a. Characters are identified by comparing them with the time-series information of the strokes (S24).

For example, as a result of performing online recognition in S24 for the input of a handwritten character “sentence”, “six” as a first candidate, “sentence” as a second candidate, and a third If the candidate "father" is generated as a candidate, the recognition result is stored in the online recognition result storage area 8b of the RAM 8 (S25).

Next, the online recognition result storage area 8b
Then, a recognition score of the online recognition result stored in is stored (S26). Specifically, a value obtained by normalizing the similarity (distance) between the recognized candidate character and the data used for recognition with a value of 0 to 100 is used. For example, in online recognition, if the time-series information of the stroke stored in the input stroke data storage area 8a and the time-series information of the stroke of the recognition candidate character completely match, the recognition score becomes “100”. If they do not match at all, the recognition score is “0”. In this way, the candidate of the online recognition result with the recognition score is, for example, “6 (6
5) "," sentence (60) ", and" father (55) "(2
6).

Next, it is determined whether or not the similarity is equal to or greater than a threshold (S27). Specifically, the recognition score obtained in S26 is stored in the primary recognition result availability determination threshold storage area 7 in the ROM 7.
It is determined whether the value is equal to or greater than a predetermined threshold value stored in h (S27). For example, if the recognition score is equal to or greater than the predetermined threshold “80” stored in the primary recognition result availability determination threshold storage area 7h (S27: YES), candidates for the integration result equal to or greater than the threshold are displayed in the display buffer of the RAM 8. Is stored in the working area 8e functioning as (S28)
It is displayed on the display section 6a of the display device 6 (S29).

Here, the candidates of the online recognition result for which the recognition score was obtained in the processing of S26 are “six (65)”, “sentence (60)”, and “father (55)”, all of which have the threshold “80”.
(S27: NO), the recognition accuracy is poor, that is, it is considered that there is no reliability.
The detailed off-line recognition is performed using the feature data (S30).

More specifically, among the characters of the candidates for the online recognition result for which the recognition score has been obtained in the processing of S26, those having a high similarity to a predetermined order are stored in the input stroke data storage area 8a. Based on the stroke information, the off-line recognition dictionary storage area 7f of the ROM 7
The detailed offline recognition is performed with reference to the 64-dimensional offline recognition dictionary stored in the storage device (S30).

Here, the CPU 2 executes the off-line recognition program stored in the off-line recognition program storage area 7d of the RAM 7 to store the information on the character shape and the input stroke data stored in the 64-dimensional off-line recognition dictionary. The character is identified by comparing it with the information on the shape of the character composed of the strokes stored in the area 8a (S30). In this detailed offline recognition, since offline recognition is performed using 64-dimensional (64) feature data, offline recognition can be performed using feature data four times larger than the coarse offline recognition described above. Is longer than coarse offline recognition, but has the characteristic that recognition accuracy is high.

When the detailed off-line recognition processing (S30) is completed, the recognition result is stored in the off-line recognition result storage area 8c (S31), and the candidate character of the recognition result stored in the off-line recognition result storage area 8c is stored in the RAM 8 It is stored in the working area 8e functioning as a display buffer (S32), and is displayed on the display unit 6a of the display device 6 as shown in FIG. 3 (S29). Then, the handwritten character recognition processing ends.

As described above, in the character recognition apparatus 1 according to the second embodiment of the present invention, only online recognition is performed first, so that the processing speed can be faster than that of the conventional hybrid recognition. . If the recognition score (similarity) of a character of a recognition candidate for online recognition is equal to or greater than a predetermined threshold, the recognition accuracy is determined to be high, and the online recognition candidate is displayed as it is, so that the speed of the recognition process is increased. However, the accuracy of the recognition result can be kept constant. In addition, if the recognition score (similarity) of the character of the recognition candidate for online recognition is less than a predetermined threshold, the recognition accuracy is determined to be low, and the detailed offline recognition process is performed again. I won't. As described above, in the character recognition device according to the above embodiment, in a character recognition device that performs character recognition in two stages, if the result of the primary recognition processing is sufficiently reliable,
Since the secondary recognition processing is not performed, the recognition speed can be increased.

It should be noted that the present invention is not limited to the above-described embodiment, and it is needless to say that the present invention can be implemented in various modes without departing from the gist of the present embodiment. For example,
In the character recognition device of the first embodiment, rough offline recognition is performed after online recognition, but online recognition and rough offline recognition may be performed in parallel. Alternatively, online recognition may be performed after coarse offline recognition. Further, in the second embodiment,
Online recognition is performed as primary recognition (S2
4, S26), if the recognition mode switch 4f of the handwritten character input tablet 4 is pressed with the pen 5, otherwise, the primary recognition of the handwritten character input to the handwritten character input tablet 4 is rough in the processing of S24. Only the offline recognition may be performed and the result may be stored (S25).

Further, there is provided a designating means of a recognition means for designating which of the online recognition, the coarse offline recognition, and the combination of the online recognition and the coarse offline recognition is to be used for the character recognition, and the recognition method designated by the designating means. Then, the handwritten character input to the handwritten character input tablet 4 may be recognized, and then the processing from S26 in the flowchart shown in FIG. 5 may be performed. Further, in the above embodiment, the judgment of the recognition accuracy in the judgment processing of S10 and the judgment processing of S27 is performed by comparing the recognition score with the threshold value. By comparing the scores of the ranking recognition candidates with a score difference of a predetermined value or more (for example, “40” or more), it may be determined that the recognition result is reliable.

[0065]

As described in detail above, in the character recognition apparatus according to the first aspect of the present invention, the similarity is determined by the similarity determination means on the recognition candidates recognized by the first character recognition means. If the degree is equal to or greater than a predetermined value, the recognition candidate is output as a recognition result.If the similarity is less than the predetermined value, the second character recognition accuracy of the handwritten character is higher than the first character recognition means. Character recognition is performed again by the character recognition means, and the recognition candidate can be output as a recognition result. Therefore, if the result of the primary recognition processing by the first character recognition means is sufficiently reliable, the secondary character recognition processing is not performed by the second character recognition means. Becomes possible.

According to the character recognition device of the second aspect of the present invention, in addition to the effect of the character recognition device of the first aspect,
The second character recognizing means, when the similarity of the recognition candidates recognized by the first character recognizing means is less than a predetermined value, the similarity among the recognition candidates output from the first character recognizing means. Character recognition can be performed again for items having a high rank to a predetermined rank, and the recognition candidate can be output as a recognition result. Therefore, the accuracy of recognition of handwritten characters is not reduced.

According to a third aspect of the present invention, in addition to the effects of the character recognition device of the first or second aspect, the second character recognition means includes the first character recognition means. As compared with, handwritten characters can be recognized using more character feature data. Therefore, by performing character recognition again by the second character recognition means, the recognition accuracy of handwritten characters can be improved.

According to a fourth aspect of the present invention, in addition to the effects of the character recognition device according to any one of the first to third aspects, the first character recognition means includes the second character recognition unit. Character recognition can be performed at a higher speed than character recognition means.

According to a fifth aspect of the present invention, in addition to the effect of the character recognition device according to any one of the first to fourth aspects, the first character recognition means includes the input of the handwritten character. Performing online recognition for recognizing handwritten characters using time-series information of constituent lines of the handwritten characters input from the means, the second character recognition means, of the handwritten characters input from the handwritten character input means Off-line recognition for recognizing handwritten characters using the entire shape can be performed. Therefore, performing on-line recognition as primary recognition can provide a character as a recognition candidate immediately after the completion of one character input, so that there is no time loss at the time of shifting to secondary recognition on-line recognition, which is effective in increasing the recognition speed.

According to a sixth aspect of the present invention, in addition to the effects of the character recognition device according to any one of the first to fifth aspects, the first character recognition means includes the handwritten character input means. The recognition result by online recognition that recognizes handwritten characters using the time-series information of the constituent lines of handwritten characters input from New recognition candidates can be generated based on the recognition result of the offline recognition for performing the recognition, and the second character recognition unit uses a larger amount of character feature data than the offline recognition performed by the first character recognition unit. Offline recognition can be performed. Therefore, in the primary recognition, coarse offline recognition is performed in which character recognition is performed with less character data than in the online recognition and offline recognition used in the second character recognition means. Recognition candidate characters that are robust and more reliable against a broken character, distortion of the entire character, and the like can be obtained by primary recognition.

In the character recognition apparatus according to the present invention, the first character recognition means recognizes the handwritten character using the time-series information of the constituent lines of the handwritten character input from the handwritten character input means. Online recognition means for performing, an off-line recognition means for recognizing handwritten characters using the whole shape of the handwritten characters input from the handwritten character input means, and a handwritten character input means by either the online recognition means or the offline recognition means. Since there is provided recognition means designating means for designating whether to perform character recognition of the input handwritten character, the recognition means designated by the recognition means designating means,
The handwritten character input from the handwritten character input means can be recognized.

According to the character recognition apparatus of the present invention, in addition to the effect of the character recognition apparatus of the present invention,
The first character recognition means is an online recognition means for recognizing a handwritten character using time-series information of constituent lines of the handwritten character input from the handwritten character input means, and an input from the handwritten character input means. At least an off-line recognition unit for recognizing a handwritten character using the entire shape of the handwritten character, and a recognition candidate generation unit for generating a new recognition candidate based on a recognition result of the on-line recognition unit and the off-line recognition unit. And a recognition means designating means for designating which of the two means performs character recognition of the handwritten character input from the handwritten character input means. The recognition means can recognize the handwritten character input from the handwritten character input means.

In the character recognition device according to the ninth aspect, in addition to the effect of the character recognition device according to the seventh or eighth aspect, the first character recognition means may include an input from the handwritten character input means. Online recognition means for recognizing handwritten characters by using the time-series information of the constituent lines of the handwritten characters, and handwritten character recognition using the entire shape of the handwritten characters input from the handwritten character input means Offline recognition means, recognition candidate generation means for generating new recognition candidates based on the recognition results of the online recognition means and the offline recognition means, the online recognition means or the offline recognition means or the online recognition means and the Designating which of the offline recognition means performs character recognition of the handwritten character input from the handwritten character input means Since a identification means specifying means, the recognition means which is specified by said recognition means specifying means, it is possible to perform the recognition of handwritten characters input from said handwriting input means.

According to the character recognition device of the present invention, in addition to the effect of the character recognition device according to any one of the first to ninth aspects, the handwriting input from the handwritten character input device by the display device can be provided. Recognition candidates recognized from characters can be displayed.

[Brief description of the drawings]

FIG. 1 is a block diagram illustrating a schematic configuration of a character recognition device 1;

FIG. 2 is a plan view of the handwritten character input tablet 4;

FIG. 3 is a plan view of the display device 6;

FIG. 4 is a flowchart of a character recognition process performed by the character recognition device 1;

FIG. 5 is a flowchart of a character recognition process performed by the character recognition device 1 according to the second embodiment;

[Description of Signs] 1 Character recognition device 2 CPU 3 I / O interface 4 Handwriting character input tablet 4a First input area 4b Second input area 4c Third input area 4d, 4e, 4f Recognition mode switch 5 Pen 6 Display device 6a Display unit 7 ROM 7a Character extraction program storage area 7b Online recognition program storage area 7c Online recognition dictionary storage area 7d Offline recognition program storage area 7e, 7f Offline recognition dictionary storage area 7g Recognition result integrated program storage area 7h Primary recognition result availability determination Threshold storage area 8 RAM 8a Input stroke data storage area 8b Online recognition result storage area 8c Offline recognition result storage area 8d Hybrid recognition result storage area 8e Working area

Claims (10)

[Claims] 1. A character recognition device for recognizing an input handwritten character and converting it into a character code, comprising: a handwritten character input means for inputting a character by handwriting; and a handwritten character input from the handwritten character input means. A first character recognizing means for converting the character code into handwritten characters; recognizing a handwritten character input from the handwritten character input means, converting the handwritten character into a character code, and rewriting the handwritten character from the first character recognizing means. A second character recognizing unit having a high recognition accuracy of the first character recognizing unit, and a similarity determining unit for determining a similarity to a basic character of a recognition candidate recognized by the first character recognizing unit. When the similarity is determined by the similarity determination unit as a result of the similarity determination by the similarity determination unit, if the similarity is equal to or greater than a predetermined value, the recognition candidate is output as a recognition result, and the similarity is determined. Constant of the case of less than a value again performs character recognition by the second character recognition means, a character recognition device and outputs the recognition candidates as a recognition result. 2. The method according to claim 1, wherein the second character recognizing means outputs from the first character recognizing means when the similarity of the recognition candidate recognized by the first character recognizing means is less than a predetermined value. 2. The character recognition apparatus according to claim 1, wherein character recognition is performed again for recognition candidates having a high similarity to a predetermined rank, and the recognition candidates are output as recognition results. 3. The method according to claim 1, wherein the second character recognition unit performs handwritten character recognition by using a large amount of character feature data as compared with the first character recognition unit. Character recognition device according to the description. 4. The character recognition apparatus according to claim 1, wherein the first character recognition unit has a higher character recognition processing speed than the second character recognition unit. 5. The method according to claim 1, wherein the first character recognition unit performs online recognition for recognizing a handwritten character using time-series information of constituent lines of the handwritten character input from the handwritten character input unit. , The second character recognition means,
The character recognition device according to claim 1, wherein off-line recognition for recognizing a handwritten character is performed using an entire shape of the handwritten character input from the handwritten character input unit. 6. A recognition result by on-line recognition for recognizing a handwritten character using time-series information of constituent lines of a handwritten character input from the handwritten character input means, Generating a new recognition candidate based on a recognition result of off-line recognition for recognizing a handwritten character using the entire shape of the handwritten character input from the handwritten character input means; The character recognition apparatus according to claim 1, wherein the means performs offline recognition using more character feature data than offline recognition performed by the first character recognition means. 7. A character recognition device for recognizing an input handwritten character and converting it into a character code, comprising: a handwritten character input unit for inputting a character by handwriting; and a handwritten character input from the handwritten character input unit. A first character recognizing unit that converts the character code into a character code, wherein the first character recognizing unit is handwritten using time-series information of constituent lines of the handwritten character input from the handwritten character input unit. Online recognition means for recognizing characters; offline recognition means for recognizing handwritten characters using the entire shape of the handwritten characters input from the handwritten character input means; and any of the online recognition means and the offline recognition means A recognition means designating means for designating whether to perform character recognition of the handwritten character input from the handwritten character input means. In recognition means designated by the designating means,
A character recognition device for recognizing handwritten characters input from the handwritten character input means. 8. An on-line recognition unit for recognizing a handwritten character using time-series information of constituent lines of a handwritten character input from the handwritten character input unit, the first character recognition unit comprising: Offline recognition means for recognizing handwritten characters using the entire shape of handwritten characters input from the input means, and recognition for generating new recognition candidates based on the recognition results of the online recognition means and the offline recognition means Recognizing means specifying at least two of the candidate generating means, the recognizing means designating which one of them performs character recognition of the handwritten character input from the handwritten character inputting means; With the recognized recognition means,
8. The character recognition device according to claim 7, wherein the handwritten character input from the handwritten character input unit is recognized. 9. An on-line recognition unit for recognizing a handwritten character using time-series information of constituent lines of a handwritten character input from the handwritten character input unit, the first character recognition unit comprising: Offline recognition means for recognizing handwritten characters using the entire shape of handwritten characters input from the input means, and recognition for generating new recognition candidates based on the recognition results of the online recognition means and the offline recognition means A candidate generation unit; and a recognition unit designation for specifying which of the online recognition unit or the offline recognition unit or the online recognition unit and the offline recognition unit performs character recognition of the handwritten character input from the handwritten character input unit. Means, the recognition means designated by the recognition means designation means,
The character recognition device according to claim 7, wherein recognition of a handwritten character input from the handwritten character input unit is performed. 10. The character recognition device according to claim 1, further comprising a display unit that displays a recognition candidate recognized from the handwritten character input from the handwritten character input unit.