JPH05189616A - On-line character recognizing device - Google Patents

Abstract

PURPOSE:To make it possible to recognize also characters written in the running hand, to prevent operability from being remarkably deteriorated due to the increment of recognizing processing caused by the increment of dictionaries and to improve a recognition rate. CONSTITUTION:A pen speed detecting part 20 (or an average stroke length detecting part 30) detects a pen speed (or average stroke length) and judges whether the pen speed (or average stroke length) is rapid writing or polite and slow writing based upon a certain threshold. Characters coincident with running-hand characters previously written in a character dictionary is applied to recognition processing with priority based upon the running-hand character degree judging result of the detecting part 20 (or 30). A parameter corresponding to the number of pictures which is previously set up in each number of pictures is corrected by the running-hand character degree. Consequently the quantity of recognizing processing can be reduced due to the reduction of useless processing and the recognition accuracy and the recognition rate can be improved.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an on-line character recognition device for identifying written characters in real time.

[0002]

2. Description of the Related Art Conventionally, there is a pattern matching method as a general character recognition method in an online character recognition apparatus. In this pattern matching method, feature points are extracted from a coordinate data string of a stroke (writing portion from pen-on to pen-off) input by handwriting. Then, the information of the extracted feature points is matched with the information of the pattern (hereinafter referred to as a registered pattern) in which the feature points are extracted and registered in advance by the same method, and character recognition is performed.

In this pattern matching method, it takes a lot of time to process which stroke of each stroke of the registered pattern each written stroke should be matched with. Also, if the balance of the whole character is disturbed,

[Outer 1] The result that the matching result is not similar is obtained, etc.
Weak to handwritten characters.

Therefore, a technique disclosed in Japanese Patent Laid-Open No. 229384/1987 is proposed as an on-line character recognition device which supplements the second point of the pattern matching method and requires a small amount of processing. In this device, large classification is performed according to the number of strokes of a writing character, and a portion to be written as a series in writing is defined as a partial pattern. Then, the medium classification is performed by the vector between the centers of gravity of the partial patterns, and the matching or the like is performed with the Q value as the characteristic parameter of the partial patterns. This makes it possible to perform character recognition that is resistant to character deformation and has a small amount of processing.

[0005]

However, the apparatus of the above-mentioned document has the following problems, although it has the advantage that it is resistant to character deformation and requires a small amount of processing. 2A to 2C are partial patterns / characters "mouth".
FIG. 3 is a diagram showing a modified example of FIG. 3, and FIG. 3 is a diagram showing a modified example of a character having a partial pattern “mouth”.

In the conventional character recognition device, the character to be recognized is limited in order to prevent deterioration of recognition accuracy.
The character to be recognized is, for example, a regular character as shown in FIGS. 2 (a) and 2 (b), or a character with several strokes connected. As shown in FIG. 2C, the continuous characters to which the strokes in each character are connected, such as in daily memo writing, are excluded.

Even if the continuous character as shown in FIG. 2C is defined as a dictionary, it can be easily recognized. However, regarding the characters of “goods” having a large number of strokes as shown in FIG. 3, considering the same continuation characters, many definitions are required from the 9 strokes of the regular print to the 3 strokes of the continuation letter. Therefore, the dictionary capacity increases, the amount of matching processing for comparing with the dictionary in the recognition process also increases with the dictionary capacity, and the recognition time increases.

In particular, since the online character recognition device is a device which recognizes the written character in real time,
If the recognition time becomes long, the operability will be greatly affected. On the other hand, with regard to the dictionary capacity, an increase in the dictionary capacity is not so problematic due to the tendency of the memory to be inexpensive and miniaturized in recent years. Rather, the ability to accurately recognize continuation characters that are written in daily life is a more important issue in terms of usability. However, it has been difficult to provide a device that has a short recognition processing time, has a good operability, has a high recognition rate, and can recognize a continuation character that is written daily.

As a problem that the above-mentioned conventional technique has, the present invention is capable of recognizing a continuous character, eliminating a remarkable deterioration in operability caused by an increase in recognition processing due to an increase in dictionary, and further improving a recognition rate. An online character recognition device that solves a difficult point is provided.

[0010]

In order to solve the above-mentioned problems, a first invention is a pre-processing unit for removing unnecessary data of a coordinate data sequence obtained by writing on a tablet and performing a linearization process. A positional relationship between the characteristic points extracted by the characteristic point extracting section and a characteristic point extracting section that extracts characteristic points representing the characteristics of strokes that form a written character from the coordinate data string linearized by the preprocessing section. An online character recognition for performing character recognition by comparing the output data of the feature point extraction section or the stroke coding section with registered pattern data registered in advance. The following measures are taken in the device.

That is, according to the first aspect of the present invention, a pen speed detection unit for detecting the pen speed based on a data string obtained by removing unnecessary data from the coordinate data string from the tablet is provided, and a threshold value in recognition processing is preset for each number of strokes. , And a weighting parameter are set, and the threshold value and the weighting parameter are corrected according to the pen speed information from the pen speed detecting unit.

In the second invention, the pen speed detecting unit is provided as in the first invention, and the degree of the continuous character which is the continuous character for each character is stored in advance as a dictionary. A part or all of the characters that do not match the output of are deleted from the candidates.

According to a third aspect of the invention, the preprocessing section of the first aspect of the invention,
An online character recognition device having a feature point extraction unit and a stroke coding unit, provided with an average stroke length detection unit that extracts an average stroke length from a feature point data string output from the feature point extraction unit, and recognizes each stroke number in advance. A threshold value and a weighting parameter in the processing are set, and the threshold value and the weighting parameter are corrected based on the average stroke length information from the average stroke length detection unit.

According to a fourth aspect of the present invention, in the online character recognition device comprising the processing section, the characteristic point extracting section, and the stroke coding section of the first aspect, the average stroke is calculated from the characteristic point data string output from the characteristic point extracting section. An average stroke length extraction / detection unit for extracting the length is provided, and the degree of the continuation character or the continuation character for each character is stored in advance as a dictionary, and a part of the characters that does not match the output from the average stroke length detection unit. Alternatively, all of them are deleted from the candidates.

According to a fifth aspect of the present invention, in the online character recognition device comprising the preprocessing section, the feature point extracting section, and the stroke coding section of the first aspect, the coordinates obtained by writing on the tablet are obtained. The pen speed is calculated from the data string after removing unnecessary data in the data string, and for each character writing,
After the calculation result is multiplied by the threshold value and the weighting parameter in the recognition process after the next step set in advance for each number of strokes or by the division by the pen speed calculation result, the threshold value and the weighting parameter are corrected. It is configured to perform character recognition processing.

According to a sixth aspect of the present invention, in the online character recognition device comprising the preprocessing section, the feature point extracting section, and the stroke coding section of the first aspect, the coordinates obtained by writing on the tablet are obtained. The pen speed is calculated from the data string after removing unnecessary data in the data string, and for each character writing,
The calculation result is discretely speed-determined by a certain threshold value, the continuation character degree stored for each character as a dictionary is referred to, the result after the speed-determination process is compared, and a part of the characters that do not match Alternatively, the processing for deleting all from the candidates is performed, and the subsequent recognition step is executed.

According to a seventh aspect, in the online character recognition device comprising the preprocessing section, the characteristic point extracting section, and the stroke encoding section of the first aspect, the characteristic point data string output from the characteristic point extracting section. The average stroke length of the strokes that make up the character is calculated, and the calculation result that is normalized by the character width is multiplied by the threshold or weighting parameter in the recognition process after the next step that has been set for each stroke beforehand, or the average. The character recognition process is performed after performing the calculation process by the division based on the stroke length calculation result and correcting the threshold value and the weighting parameter.

According to an eighth aspect of the present invention, in the online character recognition device comprising the preprocessing section, the characteristic point extracting section, and the stroke coding section of the first aspect, the characteristic point data string output from the characteristic point extracting section. The average stroke length of the strokes that make up the character is calculated, and the calculation result normalized by the character width is used to determine the average stroke length discretely by a certain threshold for each character writing, and as a dictionary for each character. Refer to the continuation character degree stored in, perform comparison processing with the result after the average stroke length determination processing, and perform the following recognition step after performing deletion processing to delete some or all of the characters that do not match from candidates. It is configured to

[0019]

According to the first aspect of the invention, since the online character recognition device is configured as described above, the pen speed detecting section detects the pen speed, and, for example, the pen speed is determined by a threshold value.
It is determined whether high-speed writing is being performed or polite slow writing is being performed, and the continuous character degree determination result is output. Then, based on the result of the continuation character degree determination, it is compared with registered pattern data that is registered in advance, for example, data regarding the continuation character that is described in advance in the character dictionary, etc. I do.
At this time, the pen speed information from the pen speed detection unit is used to correct the threshold value and the weighting parameter in the recognition process, thereby reducing the unnecessary recognition process, reducing the recognition processing amount, and further improving the recognition rate. ..

According to the second aspect of the invention, when comparing the result of the continuation character degree determination by the pen speed detecting unit and the continuation character degree previously described in the dictionary, the output from the pen speed detecting unit is matched. Delete some or all of the characters that you do not want to compare. As a result, the amount of comparison processing can be reduced.

In the third aspect of the invention, the average stroke length detection unit detects the average stroke length input by handwriting, and the average stroke length is written at high speed according to a certain threshold, or a careful slow writing is performed. Determine if you are doing. Then, by correcting the threshold value and the weighting parameter based on the average stroke length information of the average stroke length detection unit, the recognition processing amount is reduced by reducing unnecessary processing, and the recognition rate is improved. Can be achieved.

In the fourth aspect of the present invention, when comparing the result of the continuous stroke degree determination by the average stroke length detection unit and the continuous stroke degree previously described in the dictionary, the output from the average stroke length detection unit. Delete some or all of the characters that do not match with from the candidates for comparison. As a result, the amount of comparison processing can be reduced as in the second aspect of the invention.

In the fifth aspect of the invention, the pen speed is obtained by a calculation process, and the calculation result and the threshold value and the weighting parameter are calculated to correct the threshold value and the weighting parameter. Thereby, the pen speed can be easily detected, and the correction results of the threshold value and the weighting parameter can be easily obtained.

According to the sixth aspect of the invention, the speed determination can be performed with a small processing amount by performing the speed determination processing based on the threshold value which is calculated by calculating the pen speed. A comparison process is performed between the speed determination result and the pre-registered continuation character degree, and some or all of the characters that do not match are deleted from the candidates. This makes it possible to simplify comparison processing and the like. In the seventh invention, the average stroke length can be easily detected by calculating the average stroke length. By performing the correction processing of the threshold value and the weighting parameter by the calculation processing of the calculation result of the average stroke length and the threshold value and the weighting parameter, the correction processing can be simplified.

According to the eighth aspect of the invention, the average stroke length to be compared can be reduced by calculating the average stroke length and performing the determination process with a certain threshold value. Then, by comparing the average stroke length with the pre-registered continuation character degree and deleting some or all of the characters that do not match from the comparison target candidates,
The amount of recognition processing can be reduced. Therefore, the above problem can be solved.

[0026]

1 (a) and 1 (b) are functional block diagrams of an online character recognition apparatus showing an embodiment of the present invention. This on-line character recognition device is an individual circuit using an integrated circuit, or a digital signal processor (D
A tablet 1 that is configured by program control such as SP) and that inputs the position coordinates of characters with a pen touch
have. The tablet 1 includes a preprocessing unit 2, a feature point extraction unit 3, a stroke coding unit 4, a large classification unit 5, a middle classification unit 6, a partial pattern Q value matching unit 7, and a partial pattern stroke code distribution matching unit 8. Connected in order. A stroke number corresponding parameter setting unit 9 is connected to the middle classification unit 6 and the matching units 7 and 8, and the matching unit 8 is provided with an output terminal 10 to a display or the like.

Further, in the apparatus of FIG. 1A, the pen speed detecting section 20 is connected to the output side of the preprocessing section 2, and the output side of the pen speed detecting section 20 is connected to the stroke number corresponding parameter setting section 9. ing. In the device of FIG. 1B, the average stroke length detection unit 30 is connected to the output side of the feature point extraction unit 3, and the output side of the average stroke length detection unit 30 is connected to the stroke number corresponding parameter setting unit 9. ..

FIG. 4 is a flow chart showing an outline of the processing contents of the apparatus of FIG. S1 to S16 represent processing steps. In S1, the processing of the preprocessing unit 2, the pen speed detection of the pen speed detection unit 20, and the correction processing of the number-of-strokes corresponding parameter are performed. In S2, the feature point extraction process of the feature point extraction unit 3, the average stroke length detection of the average stroke length detection unit 30, and the stroke number corresponding parameter correction process are performed. S3 is a stroke coding process of the stroke coding unit 4, S
Reference numeral 4 denotes a large classification process based on the number of strokes in the large classification unit 5. S5 to S8 are processes in the middle classification unit 6, and S5 is a process for determining the end of the character dictionary. S6 is a vector calculation process between partial patterns, and the subsequent processes are performed only for those that match the continuation character degree in the character dictionary. S7 is a vector matching process between partial patterns, S8 is a matching result determination process, or a process of deleting a lower part of the character dictionary that does not match the continuation degree from the candidates. In S6 to S8,
The number-of-strokes parameter setting unit 9 sets parameters according to the number of strokes of the input character.

S9 to S12 are processes in the partial pattern Q value matching unit 7, S9 is a partial pattern Q value calculation process, S10 is a partial pattern Q value matching process, and S11.
Is a calculation process of the distance d _i , and S12 is a sorting process of the distance d _i . S13 to S16 are processes in the partial pattern stroke code distribution matching unit 8, S13 is a matching process, S14 is a partial pattern stroke code distribution calculation process, S15 is a partial pattern stroke code distribution matching distance d _s calculation process, and S16 is a sorting process.

FIGS. 5 (a) and 5 (b) are diagrams showing an example of the structure of a character dictionary provided in the apparatus of this embodiment. In this dictionary, characters can be selected by the number of strokes (stroke number). There is. FIG. 6 is a diagram showing an example of the structure of a partial pattern dictionary provided in the apparatus of this embodiment. The contents of the processes (1) to (9) of the apparatus of this embodiment will be described below.

(1) Input and preprocessing (S1) FIGS. 7A to 7C are explanatory views of the preprocessing shown in FIG. 4, and "." In the drawing represents a writing data string from the tablet or Represents a feature point. The tablet 1 of FIG. 1 is for inputting characters by writing. When characters are input by writing by the tablet 1, as shown in FIG. 7A, a writing data string {(x
_i , y _i ), i = 1, 2, ..., N _j } _j are extracted and sent to the preprocessing unit 2. The preprocessing unit 2 performs noise removal processing, moving average processing, and smoothing processing on the transmitted writing data string (S1), as shown in FIG.
As in (b), the data is linearized and output to the feature point extraction unit 3. When the continuous character is determined by the pen speed, the output of the preprocessing unit 2 is also given to the pen speed detecting unit 20.

(2) Feature Point Extraction Process (S2) The feature point extraction unit 3 performs a feature point extraction process (S2) from the output of the preprocessing unit 2. There are several methods for this feature point extraction processing. Here, as an example, a linearized data string {(x _i , y _i ), i = 1, 2, ...,
sine in the x and y directions between the data of n _j } _j (positive, negative, 0
The method of extracting the change point of the state of the signature as the characteristic point will be described. The signs XS _i and YS _i in the x and y directions between the data are obtained by XS _i = Sign (x _i −x _{i −1} ) YS _i = Sign (y _i −y _{i −1} ) (1) Expressed as +, 0,-. The signs in the x and y directions between the respective data thus obtained are compared with the signs between the previous data, and if they are the same, they are not registered as feature points,
If they are different, the state is changed and the points are registered as feature points. FIG. 7 (c) shows characteristic points to which start points and end points are added in addition to the points thus obtained. In general, this process may be referred to as linear approximation. These feature points are hereinafter referred to as segments, and the feature points are represented by {(X _i , Y _i ), i = 1,
2, ..., I _j } _j . The feature point information obtained as described above is stored in the stroke encoding unit 4
And output to the middle classification unit 6. When the continuation character degree is determined by the average stroke length, the characteristic point information is also sent to the average stroke length detection unit 30.

(3) Pen speed detection processing (S1); method of extracting pen speed from each coordinate point of the written character and discriminating the continuity of the writing character Whether the written character is written in a polite stroke, or scribbled / memo In order to determine whether it is a cursive like writing, considering the pen speed in the case of extreme strokes in which the strokes are drawn carefully and in one stroke, "one stroke writing" is more appropriate. Tends to write fast. Actual measurement of the writing time of the character "mouth" when writing in a square and when writing in a scribble shows that there is an individual difference, but it is about 1.5 to 2.5 times that when writing in a square. The writing speed is faster when scribbling. Therefore, the continuation character degree is extracted by utilizing the difference in pen speed between the regular writing and the stroke writing. Hereinafter, a method of determining the pen speed will be described.

First, the preprocessing unit 2 removes an abnormal point from the handwriting data string sent from the tablet 1 by the above-described noise removal processing. For example, when the coordinate sequence interval, which is a data sequence, is more than a certain threshold value, the data sequence after processing such as deleting the point is set to the pen speed detecting unit 2.
Then, the pen speed is determined by the pen speed detecting unit 20. The noise removal process is necessary to prevent a determination error due to an abnormal point when determining the pen speed. The noise removal process may be performed by the pen speed detection unit 20 as a noise removal process different from the process of the preprocessing unit 2. After the noise removal processing, the pen speed is determined as follows.

Writing data string {(x _i , y _i ), i = 1,
2, ..., N _j } _j (however, i; data number, j;
The stroke number, n _j ; the total number of j stroke data) is sampled and output from the tablet 1, but generally at a fixed sampling rate of 100 points / second to 200 points / second, the writing coordinates (X _i , y
_i ) is extracted. Now, set this coordinate sample time interval to Δ
Assuming Vs (sample points per second is 1 / Δt _s ), the pen speed is

[Equation 1] Can calculate the speed between each point. If there is a fixed fixed sampling rate,

[Equation 2] You can determine the pen speed with. To perform the operation of this formula, 2
The multiplication operation, the square root operation, and the like are included, and the operation time is long and the hardware configuration is complicated. Therefore, | x _{i + 1-} x
_The coordinate interval is generally approximated by _i | + | y _{i + 1} −y _i |, and the pen speed is approximated by this calculation.

FIG. 8 is a diagram for explaining the pen speed calculation method of FIG.

[0037]

[Outside 2] Now, assuming that the number of data points is n, the average pen speed V _AVE
Is Can be expressed as Since Δt _s is a constant value when the tablet sample time is fixed, The pen speed can be determined with. For example, Then,

[Equation 3] Thus, there is a method of discretely determining the pen speed.

As another method, for example, the V _AVE ^* value is directly used, or K _v = α · V _AVE ^* ... (4) α; an arbitrary fixed coefficient is set as the pen speed parameter K _v. It is used as a parameter correction for the recognition process described below. When the written character is M stroke, the average pen speed is Next to The pen speed is determined by the calculated value of.

In the case of writing, when the character frame is prepared on the tablet 1 and the size of the character frame is fixed, the above determination method may be used. However, when there is no character frame, or when the character frame changes each time, the continuation character degree cannot be accurately extracted only by the above determination method. For example, when the character frame is large, the written character of the writer becomes large, and the pen speed tends to be increased in proportion to this. On the contrary, when the character frame is small, the written character becomes small, and the pen speed tends to be slowed in proportion to this. Also, if there is no frame (this is usually
Free format), because of the personality of the writer, some write large letters, some write small letters, and the pen speed also changes in proportion to this. Therefore, a method of correcting the character size will be described.

FIG. 9 is a diagram for explaining the character width calculation from the handwritten data string in FIG. 4, and a correction method for the case of writing the character “article” will be described with reference to this diagram. As shown in FIG. 9, the minimum x _min and the maximum x _max of the x coordinate values are extracted from the writing data string of one character {(x _i , y _i ), i = 1, 2, ..., N _j } _j. To do. y
Also for the coordinates, the minimum value y _min and the maximum value y _max are calculated. Then, from the x _min , x _max , y _min , and y _max values,
The character width HX in the x direction and the character width HY in the y direction are calculated by the following equations. HX = x _max −x _min HY = y _max −y _min (6) With the character widths HX and HY, the pen speed average value V _AVE ^* is normalized by the following equation.

[0041]

[Equation 4] In this method, since division is required for each calculation of the inter-coordinate distance, normalization is performed by the product of the character widths HX and HY or an approximately added value as in the following equation. Or The pen speed average value V _AVE ^* obtained by (3)
Parameter correction or the like is performed based on the determination of the formula or the formula (4).

FIG. 10 is a functional block diagram showing a configuration example of the pen speed detecting section 20 in FIG.

The pen speed detecting unit 20 has a function of calculating the pen speed average value V _AVE ^* , and the x coordinate interval calculating unit 2
1, y-coordinate interval calculation unit 22, cumulative addition unit 23, count control unit 24, character width calculation unit 25, and cumulative addition / division unit 2
It is composed of 6. In this pen speed detection unit 20,
The handwriting data string transmitted from the tablet 1 is subjected to noise removal by the preprocessing unit 2 (x _i , y
_i ), the difference between the coordinate intervals of the writing data string (x _i , y _i ) is calculated, and the absolute value | x _{i + 1} −x _i |, | y
x-coordinate interval calculation unit 21 that calculates _{i + 1−} y _i |
And a y-coordinate interval calculation unit 22 are provided. On the output side of these calculation units 21 and 22, a cumulative addition unit 23 that adds each coordinate interval and accumulates it with the previous addition result.
Are connected.

The writing data string (x _i , y _i ) is connected to the count control unit 24, and by the count control unit 24,
The number of data n is calculated by counting each input data of each stroke.
_The number of strokes M of the character is counted by counting _j and counting for each stroke. The one-character writing data string (x _i , y _i ) is connected to the character width calculation unit 25, and the character width calculation unit 25 calculates the maximum value and the minimum value of each x, y coordinate of one character, Character widths HX and HY are calculated. The output side of the count control unit 24 and the output side of the character width calculation unit 25 are respectively connected to a cumulative addition unit 23 and a cumulative addition division unit 26, and the cumulative addition unit 23 and the cumulative addition division unit 26 perform cumulative addition. Is divided by the count value n _j , M and the character widths HX, HY to perform averaging and normalization to obtain the pen speed V _AVE ^* as an output.

Next, the operation of calculating the pen speed average value V _AVE ^* will be described with reference to FIG.

When the power is turned on and for each character written on the tablet 1, a cumulative addition unit 23, a cumulative addition and division unit 26,
And the registers in the count controller 24 are reset.
Generally, there are a character frame cutting method and a time cutting method for cutting characters. In the character frame cutout method, if there is a character frame set in advance on the tablet 1, according to the set character frame coordinates,
It is determined whether the writing data has moved from the currently written character frame to another character frame, and character cutting is performed. On the other hand, in the time cutout method, in the case of a free format in which there is no preset character frame and writing is possible in an arbitrary area of the tablet 1, time cutout, that is, after a certain time has elapsed from pen-off, the character is cut out as writing completion.

It is assumed that such character cutout determination is performed by the control of the count controller 24, and will be described below. In addition, you may provide a control part separately. When the completion of writing one character is identified by cutting out the character frame or time, the pen speed V _AVE ^*
The count control unit 24 outputs a command to reset and initialize the internal registers of the respective units, and the registers are initialized. As initial values, the inter-coordinate distance register Δl = 0 of the writing data string in the cumulative addition unit 23, the pen speed average register V _AVE ^* = 0 in the cumulative addition / division unit 26, and each stroke writing data in the count control unit 24 Each register is set in the number count register i = 1 and the stroke count register j = 1.

When a character is written on the tablet 1, the writing data string x _i , y _i from which noise has been removed by the preprocessing unit 2 is input. From the input writing data string x _i , y _i , i
The xth and the next i + 1th writing data interval is obtained by the subtractor in the x coordinate interval calculating unit 21 and the absolute value is calculated, so that | x _{i + 1} −x _i | is obtained as an output. With respect to the y-coordinates, the y-coordinate interval calculation unit 22 performs the same operation and obtains | y _{i + 1} −y _i | as an output. At this time, +1 is added to each stroke writing data number count register i of the count control unit 24. This addition is performed each time the writing data string (x _i , y _i ) is input until the end of each stroke is detected by the control unit 24.

As a method for detecting the end of the stroke, for example, a switch is provided at the tip of a pen for writing, the switch is turned on when the writing is pressed and turned off, and the stroke is started and switched off when the switch is turned on. It is common to detect the end of stroke with. This information is included in the writing data string, that is, a large / small value that is impossible as a coordinate value is included, and when this large / small value is detected, it is determined that the stroke has ended. Therefore, when the stroke end is determined, the number of data input so far, that is, the total writing data number n _{j of} each stroke is obtained. At the end of the stroke, the stroke number count register j in the count controller 24 is incremented by one. By performing the main addition until the end of one character is identified, the number M of strokes forming one character is obtained.

In order to calculate the maximum and minimum values of the x and y coordinates of one character in the character width calculation unit 25 using the input handwritten data string (x _i , y _i ), the character data is used for the first data. The x-coordinate minimum register x _min , the maximum register x _max , the y-coordinate minimum register y _min , and the maximum register y _max in the width calculation unit 25 are respectively x _min = x ₁ , x _max = x ₁ , y _mi.
The count controller 24 controls to set _n = y ₁ and y _max = y ₁ . Then, the next writing data (x _i , y
_{When i)} is input, and compared with the value of x-coordinate minimum register x _{mi n} and maximum register x _max, if less than x ₂ x
Replace _min = x ₂ with x _max = x ₂ if larger. Also, y
Comparing the value of the coordinate minimum register y _min and maximum register y _max, y _min = y ₂ if less than y _2, large if y _max =
By performing the operation of substituting with y ₂ , when the writing data for one character is input, the maximum and minimum values of the x, y coordinates of the character are stored in each register.

On the other hand, x-coordinate and y-coordinate interval calculation units 21, 2
Each time the writing data x _i , y _i is input from 2, the x coordinate, y coordinate writing data interval | x _{i + 1} −x _i |, | y _{i + 1}
−y _i | is output, and this output is cumulatively added by the cumulative addition unit 23. 1st x-coordinate of each stroke, y coordinate handwritten data interval _{| x 2 -x 1 |, |} y 2 -y 1 | is added by the adder, a register in the cumulative addition unit 23 .DELTA.l
The count controller 24 is set so that
Controlled by. Subsequent i-th x-coordinate and y-coordinate writing data intervals | x _{i + 1} −x _i |, | y _{i + 1} −y _i | are also added by the adder, and cumulatively added up to the i−1 th The value stored in the register Δl in the cumulative addition unit 23 is added, and the result is stored in the register Δl. The count control unit 24 controls such cumulative addition to be performed n _j −1 times, which is one less than the number of data until the end of each stroke is detected, that is, the total writing data number n _{j of} each stroke.

When the end of each stroke is detected, the cumulative addition unit 23
The value of the register Δl in the
This value is divided by n _j −1, which is one less than the number n _{j of} stroke writing data obtained by counting by the count control unit 24. Further, the cumulative addition / division unit 26 adds the value of the pen speed average value register V _AVE ^* (the initial value of this register V _AVE ^* is set to 0) in the cumulative addition unit 23 to add the pen speed average. Store in value register V _AVE ^* . The count control unit 24 repeats this operation until the end of writing one character is identified by the character frame cutout or the time cutout, and the pen speed average value register V _AVE ^{* of} each stroke forming one character is cumulatively added. can get.

Next, when the end of writing one character is identified, the cumulative addition value of the pen speed average value register V _AVE ^* for each stroke obtained above is divided by the value M of the stroke number count register j in the count controller 24. And average. further,
The character stored in the character width calculating section 25 x, y-coordinate maximum / minimum value register _{x min, x max, y min} , y ma x
From the value of, the character width, that is, HX = x _max −x _min , HY =
The normalized pen speed average value V _AVE ^* is output from the cumulative addition / division unit 26 by obtaining y _max −y _min by a subtracter and then dividing each by an adder, that is, an HX + HY value. Obtained as.

FIG. 11 shows a flowchart of an example of calculating the pen speed average value V _AVE ^* by the pen speed detecting unit 20 of FIG. S21 to S37 in the figure represent processing steps. First, in S21, i representing the number of each writing data and j representing the stroke number are set as initial value settings.
Is set as the initial value 1, and the pen speed average value V _AVE ^* and Δl, which is the distance between the coordinates of the writing data string, are set as the initial value 0. Also, x
In order to obtain the maximum and minimum values of the coordinate and the y coordinate, the first coordinate value of the first stroke is set as the initial value as the initial value.

Next, in S22, the inter-coordinate distances Δx and Δy of the writing data string are calculated, and these are added to obtain Δl. The calculation and cumulative addition of Δl are repeated for the number of j stroke writing data intervals n _j -1 times. Also, from S23
In S32, the minimum value and the maximum value of the x coordinate and the minimum value and the maximum value of the y coordinate for calculating the character width are calculated. I-th
X _min which is the maximum and minimum x coordinate up to the 2nd data,
The x _max value is compared with x _{i + 1} which is the x coordinate of the next data, and if it is small, it is updated to x _min , and if it is large, it is updated to x _max . By performing this calculation only n _j −1 times as the writing data interval, the maximum and minimum values of the x-coordinate of the character are calculated. The same applies to the y coordinate.

After n _j -1 times, which is one less than the strobe data number n _j , the pen speed average value V _AVE ^* for the j stroke is set to the j stroke data interval number n _j -1 in S33.
Calculate by averaging by dividing by. This calculation and cumulative addition are performed only through a few strokes through S34 and S35.

In S36 and S37, the number of strokes M
H, which is the value obtained by calculating and cumulatively adding the average pen speed value V _AVE ^* from the maximum and minimum values of the writing characters x and y, respectively.
The pen speed average value V _AVE ^* is calculated by normalizing the added value of X and HY. From this value, the pen speed is discriminated by the equation (3), or the parameters of the identification process described later are corrected by the equation (4).

(4) Average stroke length detection processing (S
2); A method to determine the continuation character degree from the average stroke length of the written character. To determine whether the written character is written in a polite style or is a cursive / scribbled cursive, a stroke is drawn. Considering the stroke length in the case of a “pencil” written carefully with one stroke and the extreme “one stroke writing”, the case of scribbled writing compared to the case where the stroke is drawn carefully with one stroke like “regular handwriting” , About 1.5 to 3 times. Therefore, the average stroke length detection unit 30 in FIG. 1B determines the continuation character degree by using the average value of the stroke lengths. Hereinafter, a method of calculating the average value of each stroke length will be described.

In calculating the average value of each stroke length, a method of using the writing data string preprocessed by the preprocessing unit 2 and a feature point extraction unit 3 for extracting a feature point described later after the feature point extraction There is a method of using the data column of. Compared to the former handwritten data string, the latter data string after feature point extraction has only some features, so there is a slight error from the stroke length of actual writing, but the number of data is small and the calculation The amount can be extremely small. Therefore, the latter example will be described.

The characteristic point data sequence after the characteristic point extraction is represented by {(X _i , Y _i ), i = 1, 2, ..., N _j } _j (where i is the characteristic point data number, j is the stroke number). Number, N
_j ; total number of feature point data of j strokes), the stroke length l _j of the j stroke is

[Equation 5] It is calculated by. Average stroke length of a single written letter l
_AVE , M is the number of writing strokes of the writing character, It is calculated by. Here, in the equation (9), since the square calculation and the square root calculation are required, the calculation time is long and the hardware configuration becomes complicated. Therefore, | X _{i + 1} −X _i |
The coordinate interval is approximated by + | Y _{i + 1} −Y _i |, and the average stroke length is calculated by the following equation (10-1). Based on the calculated average stroke length l _AVE , the continuation character degree is determined as follows. For example,

[Equation 6] In this way, the continuation character degree is discretely determined. Also,
l _AVE value is used directly, or K _l = β · l _AVE (12) β as a continuous character degree parameter K _l is set and used as parameter correction for recognition processing described below, etc. To do.

In the case of writing, when the character frame is prepared on the tablet 1 and the size of the character frame is fixed, the above determination method may be used. However, when there is no character frame, or when the character frame changes each time, the continuous character cannot be extracted accurately by the above-mentioned determination method alone. For example, when the character frame is large, the written character of the writer becomes large, and the average stroke length also becomes long in proportion to this. On the contrary, when the character frame is small, the handwritten character is small, and the average stroke length is proportionally short. Further, when there is no character frame, some writers write large characters and some write small characters, depending on the personality of the writer, and the average stroke length changes in proportion to this, correspondingly. Therefore, a method for correcting such a character size will be described.

FIG. 12 is a diagram for explaining the calculation of the character width from the feature point data string in FIG. 4, and the correction method when the character "goods" is written, for example, will be described with reference to this diagram.

As shown in FIG. 12, one-character feature point extraction data string {(X _i , Y _i ), i = 1, 2, ...
From N _j } _j , the minimum value X _min and the maximum value X _{max of the} X coordinate value
Is extracted, and the minimum value Y _min and the maximum value Y _max of the Y coordinate are extracted. _{Then, X min, X ma x,} Y min, the Y _max value, the character portion HX and y direction of the x-direction character width HY calculated by the following equation. _{HX = X max -X min HY =} Y max -Y min ··· (13) The character width HX, the HY, the average stroke length obtained in (10-1) below, normalized by the following equation ..

[0064]

[Equation 7] Alternatively, since division is required for each calculation of the inter-coordinate distance, normalization is performed by the product of the character widths HX and HY, or approximately by the added value. in this case, Or Parameter correction or the like is performed from the average stroke length l _AVE obtained in step 1 by the determination of equation (11) or equation (4).

FIG. 13 is a functional block diagram showing a configuration example of the average stroke length detection unit 30 in FIG. The average stroke length detection unit 30 has a function of calculating the average stroke length l _AVE , and has an x coordinate interval calculation unit 31, ay coordinate interval calculation unit 32, a cumulative addition unit 33, a count control unit 34,
It is composed of a character width calculation unit 35 and a division unit 36.

In the average stroke length detecting section 30, the characteristic point data string (X _i ,
Y _i ), and the feature point extraction data string (X _i , Y
_i ) The coordinate interval difference is calculated, and the absolute value | X _{i + 1} −X
x-coordinate interval calculation unit 3 for calculating _i |, | Y _{i + 1} −Y _i |
A 1 and y coordinate interval calculator 32 is provided. On the output side of these calculation units 31 and 32, a cumulative addition unit 33 that adds each coordinate interval and accumulates it with the previous addition result.
Are connected.

The feature point data string (X _i , Y _i ) is connected to the count control unit 34, and the count control unit 34 counts each input data of each stroke to count the number of data N _j , and further strokes. The number of strokes M of each character is counted. The character point data sequence X _i , Y _i of one character is connected to the character width calculation unit 35, and the character width calculation unit 35 calculates the maximum and minimum values of the x and y coordinates of one character to obtain the character width HX. , HY is calculated. Count control unit 34 output side and character width calculation unit 35
The output side of is connected to the cumulative calculation unit 33 and the division unit 36, respectively, and by the cumulative addition unit 33 and the division unit 36,
The result of cumulative addition is the count value M and the character widths HX and HY.
Then, the average stroke length l _AVE is obtained as an output.

Next, the operation of calculating the average stroke length l _AVE will be described with reference to FIG.

The registers in the cumulative addition unit 33, the division unit 36, and the count control unit 34 are reset when the power is turned on and for each character written on the tablet 1. As described above, the character cutout generally includes the character frame cutout method and the time cutout method. Regarding the character frame cutting method, when there is a character frame set in advance on the tablet 1, it is determined whether or not the writing data has moved from the currently written character frame to another character frame by the set character frame coordinates, Cut out characters. on the other hand,
In the time cutout method, in the case of a free format in which there is no preset character frame and writing is possible in an arbitrary area of the tablet 1, time cutout, that is, a character is cut out as writing completion after a certain time has elapsed from pen-off.

It is assumed that such character cutout determination is performed by the control of the count controller 34, and will be described below. In addition, you may provide a control part separately. When the completion of writing one character is identified by character frame cutout or time cutout, the calculation result of the average stroke length l _AVE is output, and thereafter, the internal register of each unit is reset, and the count control unit 34 instructs the initialization. Is output and the register is initialized.
As the initial value, the inter-coordinate distance register Δl = 0 of the feature point data string in the cumulative addition unit 33, each stroke feature point data number count register i = 1 in the count control unit 34,
Each register is set in the stroke count register j = 1.

When the characteristic point data string (X _i , Y _i ) is input from the characteristic point extracting unit 3, the i-th and the next i + 1-th characteristic point are extracted by the characteristic point data string (X _i , Y _i ). The data interval is obtained by the subtractor in the x-coordinate interval calculation unit 31, and the absolute value is calculated to obtain | X _{i + 1} −X _i | as an output. With respect to the y-coordinates, the y-coordinate interval calculation unit 32 performs the same operation and obtains | Y _{i + 1} −Y _i | as an output. At this time, +1 is added to each stroke feature point data number count register i of the count control unit 34. This addition is performed by the control unit 34 at the end of each stroke.
It is performed every time the feature point data string (X _i , Y _i ) is input until it is detected by.

As a method of detecting the end of the stroke, as described above, for example, a switch is provided on the pen tip for writing, and the switch is turned on when the writing is pressed,
Generally, the switch is turned off when separated, and the stroke start is detected when the switch is turned on, and the stroke end is detected when the switch is turned off. Including this information in the feature point data string,
That is, an impossible large / small value is included as a coordinate value, and when this large / small value is detected, it is determined that the stroke has ended. Therefore, when the stroke end is judged, the number of data input so far, that is, the number N _j of feature point data of each stroke is obtained. At the end of the stroke, the count controller 3
The stroke count register j in 4 is incremented by +1. By performing the main addition until the end of one character is identified, the number M of strokes forming one character is obtained.

Input feature point data string (X _i , Y _i )
Therefore, in order to calculate the maximum and minimum values of the x and y coordinates of one character in the character width calculation unit 35, in the case of the first data, the x coordinate minimum register X _min and maximum register X in the character width calculation unit 35 are calculated. _max , y-coordinate minimum register Y _min , and maximum register Y _max are respectively set as X _min = X ₁ , X _max = X ₁ , Y
_The count controller 34 controls to set _min = Y ₁ and Y _max = Y ₁ . And the next writing data (X ₂ ,
Y ₂ ) is input, the values of the x-coordinate minimum register X _min and maximum coordinate X _max are compared, and if smaller than X ₂ , X
Replace with _min = X ₂ , and if larger, X _max = X ₂ . Also, y
Comparing the value of the coordinate minimum register Y _min and the maximum register Y _max, if less than Y ₂ Y _min = Y _2, large if Y _max =
By performing the operation of substituting with Y ₂ , when the writing data for one character is input, the maximum value and the minimum value of the x, y coordinates of the character are stored in each register.

On the other hand, x-coordinate and y-coordinate interval calculation units 31, 3
Every time the feature point data X _i , Y _i is input from 2, the x coordinate, y coordinate feature point data interval | X _{i + 1} −X _i |, | Y
_{i + 1−} Y _i | is output, and this output is cumulatively added by the cumulative addition unit 33. The first x coordinate of each stroke,
y coordinate feature point data interval _{| X 2 -X 1 |, |} Y 2 -Y 1
| Is added by the adder and set in the register Δl in the cumulative addition unit 33 as an initial value. Subsequent i-th x-coordinate and y-coordinate feature point data interval | X _{i + 1} −X _i |, |
Y _{i + 1} −Y _i | is also added by the adder, and i−
It is cumulatively added up to the first, is added to the value stored in the register Δl in the cumulative addition unit 33, and is stored in the register Δl. The count controller 34 controls such cumulative addition to be performed N _j −1 times, which is one less than the number of data until the end of each stroke is detected, that is, the total number N _j of feature point data of each stroke.

When the character cutout is detected, the value of the register Δl in the cumulative addition unit 33 is sent to the division unit 36, and this value is stored in the value M of the stroke number count register j in the count control unit 34.
Divide by and average. Further, the x / y coordinate maximum / minimum value register X of one character stored in the character width calculation unit 35.
_min, X _max, obtained Y _min, than the value of Y _max, character width, i.e. HX = X _max -X _min, by the HY = Y _max -Y _min subtractor. Next, the value obtained by adding each with an adder, that is, H
By dividing by the X + HY value, the normalized l _AVE is obtained as the output of the division unit 36.

FIG. 14 shows a flow chart of an example of calculating the average stroke length l _AVE by the average stroke length detection unit 30 of FIG. S41 to S57 in the figure represent each processing step. First, in S41, as the initial value setting, i representing the number of the feature extraction data of each stroke and j representing the stroke number are set to the initial value 1, and the average stroke length l _AVE and the inter-coordinate distance of the feature point data string are set. There is Δl
Is an initial value 0. Further, in order to obtain the maximum and minimum values of the x coordinate and the y coordinate, the first value of the first stroke is set as the initial value.
Set the coordinate value as the initial value.

Next, in S42, the inter-coordinate distances ΔX and ΔY of the feature point data string are calculated, and these are added to obtain Δl. The calculation and cumulative addition of Δl are repeated N _j −1 times for the characteristic points of the j stroke. Further, in S43 to S52, the maximum value and the minimum value of the x coordinate and the maximum value and the minimum value of the y coordinate for calculating the character width are calculated. The x-coordinate maximum and minimum values X _min and X _max up to the i-th data are compared with the x-coordinate X _{i + 1 of} the next feature point, and X _{min is} small and X _{max is} large. To update. By performing this calculation only N _j −1 times at feature point data intervals, the maximum and minimum values of the x-coordinate of the character are calculated. The same operation is performed for the y coordinate to obtain the maximum and minimum values.

In S56 and S57, the number of strokes M
H calculated by calculating the average stroke length l _AVE and accumulatively adding it from the maximum and minimum values of x and y of the written character, respectively.
The average stroke length l _AVE is calculated by normalizing with the added value of X and HY. From this value, the continuation character degree is discriminated by the equation (11), or the parameter correction of the recognition processing described later is conducted by the equation (12).

(5) Stroke Encoding Process (S3) The stroke encoding unit 4 in FIG. 1 encodes each stroke based on the characteristic point information obtained by the characteristic point extracting unit 3. There are many methods for this encoding, but generally, the encoding is performed by classifying by the X and Y signs of each segment, the angle of the segment, and the rotation angle between the segments.

FIG. 15 is an explanatory diagram of this encoding process.
θ ₁ , θ ₂ , and θ ₃ indicate the angles of the segments (angles formed with the + x direction), and θ ₁ ⁻ and θ ₂ ⁻ indicate the rotation angles between the adjacent segments. The coded straw data is output to the large classification unit 5 and the partial pattern stroke code distribution matching unit 8.

(6) Major classification processing (S4) The major classification unit 5 in FIG. 1 receives the output of the stroke coding unit 4 and performs major classification on the target character according to the number of strokes. Therefore, a character that can be the number of strokes for each number of strokes (stroke number) is prepared in advance in a character dictionary as shown in FIGS. As shown in FIG. 2 and FIG. 3, the dictionary of FIG. 5 is created in consideration of the case where strokes are connected as a modification and a continuous character is formed. For example, "word" is the character of 14 strokes when written in a regular script,

[Outside 3] 4 strokes each 3 strokes 3 strokes 2 /
It is created with consideration of a few consecutive characters that make up one stroke and a total of 10 strokes. In addition, "wish" is the character of 19 strokes when written in a regular script,

[Outside 4] Extreme stroke characters such as 2 strokes each having 1 stroke, 5 strokes having 2 strokes, 6 strokes having 2 strokes and 3 strokes having 2 strokes are also taken into consideration and created. For example, it is assumed that the stroke number of the character pattern input by handwriting is 10 strokes. In this case, among the characters stored in the character dictionary, the character “mute” that can be 10 strokes as shown in FIG.
"Dust", "Ai" ... are selected as candidate characters.

(7) Medium Classification Processing (S5 to S8) In the medium classification section 6 of FIG. 1, candidate characters obtained by large classification in the large classification section 5 according to the number of strokes are converted into partial pattern vectors described below. Further classify. Here, the partial pattern refers to a portion of one character that is written as a series in writing, and the inter-partial-pattern vector is the center of gravity of one partial pattern and the center of gravity of another partial pattern. , The vector to be the end point.

First, an example of a method of calculating a vector between partial patterns will be described. X, y of each segment in the partial pattern
Letting the components be (dx _i , dy _i ), the length dl _i of each segment is

[Equation 8] It is represented by. In addition, the center coordinates of each segment normalized by dividing by the character widths HX and HY are (x _i ^* , y
_i ^* ), the barycentric coordinates (X _w ,
Y _w ) is

[Equation 9] Required by. The center of gravity of each partial pattern is calculated by the above method, and the vector between partial patterns is calculated with the center of gravity of one partial pattern and the center of gravity of another partial pattern as the starting point and the ending point, respectively. Note that here, the inter-partition-pattern vector is considered in the x direction and the y direction, respectively.

As an explanatory view of the vector between the partial patterns,
In Figure 16,

[Outside 5] In the middle classifying unit 6, by the vector between the partial patterns,
The candidate characters selected by the large classification unit 5 are narrowed down to perform middle classification. Here, as an example, consider the case where "O" is written and input, and the procedure of middle classification for this input character will be described below. ..

Since the character "O" written and input has 10 strokes, the 10 stroke portion of the character dictionary shown in FIG. 5 is referred to. Then, the word "mute" is arranged in the first place here,
In that column, the partial patterns that make up the "voice", the writing order of the partial patterns, the stroke number information of each partial pattern (hereinafter referred to as the cut position), and the vector value between partial patterns calculated in advance from the registered pattern are shown. Has been done.
Similar information is lined up for the characters "dust" and "a" in the following order, and the middle classification unit 6 determines whether or not to make a candidate according to this character order, and performs middle classification as follows.

First, assuming that the handwritten input character is a "mute", the matching distance d _vec of the partial pattern vector
Ask for. As written in the character dictionary in Figure 5,
“Muta” has a cutting position of (3, 7),

[Outside 6] In this example, since the input pattern is "a", considering the inter-pattern vector for "a" at this cut position, it becomes as shown in FIG.

[0087]

[Outside 7] The matching distance d _vec is calculated by the following equation.

[0088]

[Equation 10] Generally, when the number of partial patterns of written characters is multiple,
The normalization is performed by the number of partial patterns BPN, and the matching distance d _vec is

[Equation 11] Is calculated according to. Here, a certain threshold value VECREJ is set, and it is determined whether or not the calculated d _vec is larger than VECREJ. And when d _vec > VECREJ,
Assuming that the character is not the referenced character (in this case, "mute"), the partial pattern vector of the next character is matched. d
_{When vec} ≤ VECREJ, it is considered that it is a "mute", and the calculation and matching of the partial pattern Q value described below are performed. The threshold value VECREJ is set in advance in the stroke number corresponding parameter setting unit 9 for each stroke number, and a value is set according to the stroke number of the input pattern at the time of recognition.

(7-1) Correction during classification processing; method of using pen speed information or average stroke length information as parameter correction for recognition processing, etc. When a character has a small number of strokes, the amount of information in the partial pattern vector is small. Therefore, in this case, the ranked ranks have uncertainty, and the threshold value VECREJ needs to be large. On the contrary, in the case of a character having a large number of strokes, it is preferable to reduce the threshold value VECREJ for the same reason. However, like stroke writing, when the strokes are connected and the characters are deformed, and when writing with a regular script, even characters that have a large number of strokes and have sufficient partial pattern vector information,
The number of strokes decreases. Therefore, the threshold value VECR previously stored in the stroke number corresponding parameter setting unit 9 corresponding to this is set.
The EJ becomes large, candidates are not narrowed down, many candidates remain, the processing amount of the next step increases, and the partial pattern vector information is not fully utilized. Therefore,
It is necessary to correct the threshold value VECREJ by the continuous character, and the method will be described below.

As described above, when the continuation character degree is high, the number of strokes tends to decrease from the original number of strokes, so the threshold value VECREJ is corrected in the direction of the original number of strokes. Since a character having a large number of strokes has sufficient partial pattern vector information and is highly deterministic, it is preferable to lower the threshold VECREJ. Therefore, the threshold VECRE becomes
It suffices to correct J downward. Therefore, as the output of the pen speed detection unit 20 or the average stroke length detection unit 30, the formula to be corrected by the pen speed parameter K _{v of} the formula (4) and the continuous character parameter K _l of the formula (12) is as follows. Like VECREJ = VECREJ / K _v or VECREJ = VECREJ / K _l Due to this correction, when the pen speed is fast (K _{v is} large), the continuation degree is large and the number of strokes tends to decrease.
It is corrected to lower EJ. Further, when the average stroke length is long (K _{1 is} large), the continuation character degree is large and the number of strokes tends to decrease, so the threshold value VECREJ is corrected to a lower value. From the same meaning, it can be corrected by subtraction. In this case, VECREJ = _{VECREJ-K v} or VEC
It is corrected by REJ-K ₁ .

(7-2) Narrowing down during middle classification processing; method of describing continued character degree in dictionary and narrowing down candidates As shown in FIG. 5 (b), continued character degree of each character previously defined in the character dictionary Is described. The continuation character degree in this dictionary is set to the continuation character degree of 1 when 10 strokes are drawn when writing carefully on a regular script, for example, “mute”.

[0092]

[Outside 8] A character whose number of strokes is drastically reduced compared to the number of strokes written in a regular drawing and has 10 strokes has a continuation character degree of 3.

[0093]

[Outside 9] Characters that have 10 strokes due to some continuation characters are described in the dictionary in advance as a continuation character degree 2.

From the pen speed detecting unit 20 or the average stroke length detecting unit 30, the pen speed is low as shown in the equation (3),
As the medium speed and high speed information, information having a short average stroke length and a long average stroke length is transmitted as in the equation (11), and candidates are narrowed down based on this information. By narrowing down the candidates, it is possible to reduce unnecessary processing after the next step, reduce the processing amount, and improve the recognition processing speed.

For example, from the pen speed detection unit 20, (3)
It is assumed that the pen speed information is low speed 1 according to the formula. As described above, the correlation between the pen speed and the degree of continuous writing has a relationship that the pen speed of the continuous writing is high, and the pen speed is slow in the case of careful writing such as regular writing. When the pen speed is 1 and is low as in this example, it is determined that the writing is done in a polite manner, and the processing with the continuous character degree of 1 in the character dictionary is given priority.

For example, from the candidates obtained by the partial pattern vector matching, the ones with the continuation character degrees 2 and 3 are deleted from the partial candidates. A part of the lower part of the partial pattern matching result is deleted from the candidates. Alternatively, the partial pattern vector matching process is not performed on the characters having the continuation character degrees 2 and 3, and the characters having the continuation character degrees 2 and 3 are not necessarily candidates for the processing. In this case, the continuity is small,

[Outside 10] Further, let us consider a case where the average stroke length detection unit 30 has transmitted by the equation (11) that the average stroke length is 3 and is long, and arrives. When the average stroke length is long, it is determined that the character is greatly deformed like stroke writing and the number of strokes due to the connection of strokes is greatly reduced. Then, the character dictionary with the continuous character degree of 3 is preferentially processed. For example, from the candidates obtained by the partial pattern vector matching, the ones with the continuous character degrees 1 and 2 are partially deleted.
A part of the lower part of the partial pattern vector matching result is deleted from the candidates. Alternatively, the partial pattern matching process is not performed on the characters having the continuation character degrees 1 and 2, and the characters having the continuation character degrees 1 and 2 are not necessarily candidates for the processing. In this case, the continuation is high,

[Outside 11] (8) Partial Pattern Q-Value Matching Processing (S9) The partial pattern Q-value matching unit 7 of FIG. 1 calculates the partial pattern Q-values of the candidate characters remaining in the intermediate classification by the partial pattern vector in the intermediate classification unit 6, Matching is performed with the partial pattern Q value in the partial pattern dictionary shown in FIG. The partial pattern Q value of this partial pattern dictionary is created in advance from the registered pattern and stored. Here, the partial pattern Q value refers to a characteristic parameter indicating the length, direction, and position of each segment.
In online character recognition, the movement of the writing pen is
The X, Y direction, + or-direction is obtained as important information, and it is this partial pattern Q that effectively uses this information.
It is a value.

First, the method of calculating the partial pattern Q value will be described. In the following equations (21) to (28), Σ represents all strokes, addition regarding all segments, and HX and HY represent character widths.

[0098]

[Equation 12]

[Equation 13] In the case of the expressions (21) to (28), the values are the values of the respective directional positions when the origin is set to the lower left. At this time, the values near the origin become 0 when subjected to multiplication. for that reason,
In order to prevent it from becoming 0, the origins are exchanged, and the values Q _{9 to} Q ₁₆ at each direction position when the origin is set to the upper right are described in the same manner, and by the total of 16 values of Q _{1 to} Q ₁₆ , It shall represent the length, direction and position of each stroke segment of the target character. The partial pattern Q value matching unit 7 calculates the partial pattern Q value with respect to those remaining by the classification by the partial pattern vector. For example, “A” is written and “dust” is generated between the partial patterns. It is assumed that they remain after being classified by the vector. in this case,
As shown in Fig. 5, the cut position of "dust" is (3, 2,
5) and

[Outside 12] Cut the input pattern at the cutting position (3, 2, 5),
Calculate Q ₁ ^{* to} Q ₁₆ ^* respectively.

[0099]

[Outside 13] The partial pattern Q values Q ₁ ^{* to} Q ₁₆ ^* calculated respectively are shown in FIG.
Matches with the Q value of the partial pattern in the partial pattern dictionary.

[0100]

[Outside 14] The sum of the differences in these matchings is the matching distance d _BP . At this time, the distance d _BP represents how close the input pattern “Ai” is to “Dust”. In general,
According to the stroke number BS _j of each partial pattern, the following equation (2
9) Weighting is performed and the matching distance d
_{Let's call} it _BP .

[0101]

[Equation 14] Further, the weighting parameters w _vec and w _BP of the following equation (30) are determined in advance for each number of strokes and stored in the number-of-strokes corresponding parameter setting unit 9, and set by the number-of-strokes parameter setting unit 9 according to the number of strokes of the input pattern at the time of recognition. Weighting is performed according to the weighted parameter value. Then, the distance d _BP obtained in this way and the d _vec obtained by matching the partial pattern vectors obtained in the previous step are weighted by w _vec and W _BP , respectively, to obtain the distance d _i . .. d _i = w _{vecd vec} + w _{BPd BP} (30) The above operation is performed for all the candidate characters remaining in the classification by the inter-partition pattern vector, and the sorting by d _i is performed.

(9) Partial Pattern Stroke Code Distribution Matching Processing (S13 to S16) The partial pattern stroke code distribution matching unit 8 of FIG. 1 receives the outputs of the partial pattern Q value matching unit 7 and the stroke coding unit 4, A partial pattern stroke code distribution is obtained for candidate characters narrowed down by classification. Then, this distribution is matched with the partial pattern stroke code distribution created in advance from the registered pattern and stored in the partial pattern dictionary in FIG. 6, and the higher-ranked candidates are ranked.

The range of objects to be ranked is determined by the ratio with the distance d ₁ of the first candidate obtained by sorting d _i , for example. That is, ranking is performed by setting the candidate range of d _j / d ₁ ≤ZRATE as the target range. Here, the threshold value ZRATE is set in advance in the stroke number corresponding parameter setting unit 9 for each stroke number of the written character, and at the time of recognition, a value is set according to the stroke number of the input pattern.

For a character having a small number of strokes, the amount of information in the partial pattern vector and the partial pattern Q value is small. Therefore, in this case, the ranks ranked by d _{i in the} equation (30) tend to be indefinite, and it is necessary to take a large threshold ZRATE of the target range of the partial pattern stroke code distribution matching unit 8. On the contrary, in the case of a character having a large number of strokes, it is preferable to reduce the threshold value ZRATE for the same reason. With such a tendency, the number of strokes is stored in the stroke number corresponding parameter setting unit 9.

However, like stroke writing, when the strokes are connected and the characters are deformed, and when originally written in a regular drawing, a character with a large number of strokes has a sufficient amount of information in the partial pattern vector and the partial pattern Q value. The letters
The number of strokes decreases. Then, the threshold value ZRATE stored in advance in the stroke number corresponding parameter setting unit 9 corresponding to this is set.
Becomes larger, the target range of the partial pattern stroke distribution matching unit 8 becomes wider, and the lower patterns are also targeted.
Therefore, the sorting processing amount is unnecessarily increased, and the partial pattern vector information and the partial pattern Q value information are not fully utilized. Therefore, it is necessary to correct the threshold value ZRATE by the continuous character, and the correction method will be described.

As described above, when the continuation character degree is large, the number of strokes tends to decrease from the original number of strokes, so the threshold ZRATE is corrected in the direction of the original number of strokes. A character having a large number of strokes has a sufficient partial pattern vector and partial pattern Q value information and is highly deterministic. Therefore, it is preferable to lower the threshold ZRATE and narrow the resorting range. Therefore, the threshold ZRATE may be corrected in the lower direction as the continuation character degree increases.

From the above, as the output of the pen speed detecting unit 20 or the average stroke length detecting unit 30, the formula for correcting the pen speed parameter K _{v of} the formula (4) or the continuous character degree parameter K _l of the formula (12) is used. Is as follows. ZRATE = ZRATE / K _v or ZRATE /
K _l Due to this correction, when the pen speed is fast (K _{v is} large), the continuous character degree is large and the number of strokes tends to decrease.
It is corrected to lower TE. Further, when the average stroke length is long (K ₁ large), the continuation character degree is large and the number of strokes tends to decrease. Therefore, the threshold ZRATE is corrected to a lower value. From the same meaning, it can be corrected by subtraction. In this case, ZRATE = ZRATE-K _v or ZRATE
-Corrected by K _l .

Next, a method of calculating the partial pattern stroke code distribution will be described. As an example, it is assumed that the input pattern is “A” and the character selected as the first candidate is “A”. From the character dictionary of FIG. 5, the cut position of the candidate character "Ai" is (3, 4, 3),

[Outside 15] The input pattern "Ai" is cut at this position. In this case, the partial pattern obtained by cutting is the same as the content of the character dictionary, but the distribution of the number of stroke codes obtained by the stroke encoding unit 4 is calculated for each partial pattern.

[0109]

[Outside 16] The stroke code distributions of "01" for one, "03" for one, and "05" for one are obtained. The partial pattern stroke code distribution calculated in this way is calculated by a similar procedure from several registered patterns in advance and averaged to create the partial pattern stroke code distribution of the partial pattern dictionary of FIG. To be matched.

[0110]

[Outside 17] That is, the input pattern partial pattern dictionary "01" ... 1 book "01" ... 0.9 book "02" ... 0 book "02" ... 0.1 book "03" ... 1 Book “03” ・ ・ ・ 0.4 book “04” ・ ・ ・ 0 book “04” ・ ・ ・ 0.6 book “05” ・ ・ ・ 1 book “05” ・ ・ ・ 1 book

[Outside 18] Then, these are used as the number of partial pattern stroke codes B
Normalized by sj, the sum of the distances d _s of the matching distances of the partial patterns that are normalized, calculated from the following equation (31).

[0111]

[Equation 15] Next, the weighting parameters w _vec , w _BP , and w _{s are} calculated with respect to the distance d _vec obtained by the vector matching between the partial patterns, the distance d _{BP obtained} by the partial pattern Q-value matching, and the distance d _{s obtained} by the partial pattern stroke code distribution matching obtained above. Thus, the distance value D is calculated by the following equation (32). Weighting parameters w _vec , w _BP , w _s
Is a parameter stored in the stroke number corresponding parameter setting unit 9 after the optimum value of the weighting parameter is calculated in advance for each stroke number. The weighting parameter is set by the stroke number corresponding parameter setting unit 9 according to the stroke number of the input pattern. It is a value. D = w _vec · d _vec + w _BP · d _BP + w _s · d _s (32) This distance D is calculated for each candidate character of d _j / d ₁ ≦ ZRATE, and the candidate character is calculated according to the obtained distance D. Ranking is performed, and the recognition result is output from the output terminal 10 to a display device or the like (not shown).

Here, in the case of a character having a small number of strokes, the amount of information in the partial pattern vector and the partial pattern Q value is small. Therefore, the weighting w _{vec of} the equation (32),
w _BP should be small. Also, w _s, when a small number of strokes, is in sufficient amount of information for the stroke coding is therefore should be greater. On the contrary, in the case of a character having a large number of strokes, a vector between partial patterns and a partial pattern Q
The amount of information in the value is large, but the size of each stroke is small. Therefore, w _vec and w _BP should be large and w _s should be small because the amount of information for stroke coding is small.

As described above, the weights w _vec , w _BP , and w _s in the equation (32) are changed for each stroke number and preset in the stroke number corresponding parameter setting unit 9. However, like stroke writing, when a stroke is connected and the character is deformed, and when originally written in a regular writing, it is a character with a large number of strokes and has a sufficient amount of information in the partial pattern vector and the partial pattern Q value. As for characters, the number of strokes also decreases. Then, the weighting parameters w _vec and w _BP stored in advance in the stroke number corresponding parameter setting unit 9 corresponding thereto become large, and w _s becomes small, so that the weighting deviation occurs. Therefore, the weighting parameters w _vec , w _BP ,
It is necessary to correct w _s , which will be described below.

As described above, when the continuation character degree is large,
Originally, since the number of strokes tends to decrease, the weighting parameters w _vec , w _BP , and w _s are corrected in the direction of the larger number of strokes. Characters with a large number of strokes have sufficient partial pattern vector and partial pattern Q-value information, and are highly deterministic.
Increase w _vec and w _BP . Stroke-encoded information is less deterministic because each stroke is smaller and less deterministic, so it is better to lower w _s . Therefore, w _vec and w _BP may be increased and w _s may be decreased in accordance with the continuity of the character.

From the above, as the output of the pen speed detecting unit 20 or the average stroke length detecting unit 30, the formula for correcting the pen speed parameter K _{v of} the formula (4) and the continuous character degree parameter K _l of the formula (12) is , Is as follows. W _VEC = W _VEC · K _v or W _VEC · K _l W _BP = W _BP · K _v or W _BP · K _l W _s = W _s / K _v or W _s / K _l Pen speed is high due to this correction In the case (K _v large), the continuation is large and the number of strokes is decreasing, so W _VEC , W
It is corrected so that _BP is large and W _s is small. Further, when the average stroke length is long (K ₁ large), the continuous character degree is large and the number of strokes tends to decrease. Therefore, W _VEC and W _BP are large and W is large.
It is corrected to make _s small.

From the same meaning, as a simple method, correction can be performed by addition and subtraction. In this case, W _VEC = W _VEC + K _v or W _VEC + K _l W _BP = W _BP + K _v or W W _BP + K _l W _s = W _s −K _v or W _s −K _l is corrected. Therefore, the shift in weighting due to the continuous letters is corrected, and the recognition rate can be improved.

The present invention is not limited to the above-described embodiment, and for example, the pen speed detecting unit 20 and the average stroke length detecting unit 30 in FIG. 1 may be configured by functional blocks other than those in FIGS. 10 and 13, or Various modifications can be made such that the contents of the other blocks in FIG.

[0118]

As described above in detail, according to the first aspect of the invention, since the pen speed detecting portion is provided, even if a character is deformed due to a stroke like a stroke, the stroke is connected. It is possible to extract pen speed information such as the continuation character degree which is the frequency. Then, the output data of the feature point extraction unit or the stroke coding unit is compared with the registered pattern data that is registered in advance, for example, the continuation character degree previously described in the character dictionary, and the continuation character degree extracted by the pen speed detection is compared. The character recognition process can be preferentially performed on the characters that match the above, and the wasteful process amount can be reduced, the recognition process amount can be reduced, and the recognition process can be speeded up. Further, the stroke number corresponding parameter consisting of a threshold value and a weighting parameter set in advance so that each parameter of the recognition process becomes a value corresponding to each stroke number is corrected by, for example, the continuous character degree extracted by the pen speed detection. ing. Therefore, the fluctuation in the number of strokes due to the continuation characters is corrected
The character recognition accuracy and its recognition rate can be improved.

According to the second aspect of the present invention, the continuation degree described in the dictionary is referred to, and some or all of the characters that do not match the output from the pen speed detection unit are deleted from the comparison target candidates. By reducing the amount of comparison processing, the amount of recognition processing can be reduced, and high-speed recognition is possible.

According to the third aspect of the present invention, since the average stroke length detection unit is provided, extraction of the continuous character degree, which is the connection frequency of strokes, is possible even for a character in which strokes such as stroke writing are connected and character deformation occurs. You can do it. Then, based on the average stroke length information from the average stroke length detection unit, the stroke number corresponding parameters such as the preset threshold value and weighting parameter are corrected.
Similarly to the invention described above, for example, based on the continuation character degree extracted by the average stroke length detection, the stroke number variation due to the continuation character can be corrected, and the recognition accuracy and the recognition rate can be improved.

According to the fourth invention, since the average stroke length detection unit is provided and some or all of the characters that do not match the output thereof are deleted from the comparison target candidates, the recognition processing is performed by reducing the comparison processing amount. The amount can be reduced and the recognition processing can be speeded up.

According to the fifth invention, the pen speed is obtained by calculation, and the calculation result and the threshold value and the weighting parameter are processed to correct the threshold value and the weighting parameter. And the correction process is simple and easy.

According to the sixth invention, since the pen speed is arithmetically processed and the speed judgment processing is performed on the arithmetic result by a certain threshold value, the pen speed information to be compared is reduced and the comparison processing is performed. Will be easier. Furthermore, since a part or all of the characters that do not match are deleted from the candidates by performing comparison processing with the result after the speed determination processing, the recognition processing amount can be reduced and the recognition accuracy and recognition rate can be improved.

According to the seventh aspect of the invention, the average stroke length is calculated, the threshold value and the weighting parameter are calculated, and the threshold value and the weighting parameter are corrected. In addition to being obtained, the recognition accuracy and recognition rate can be improved.

According to the eighth aspect of the invention, the average stroke length is calculated, and the average stroke length is discriminated discretely by a certain threshold value with respect to the calculation result. Therefore, the average stroke length can be easily obtained. At the same time, the number of average stroke lengths to be compared is reduced, and the amount of comparison processing can be reduced. Furthermore, since some or all of the characters that do not match are deleted from the candidates, the recognition processing amount can be reduced and the recognition rate and recognition accuracy can be improved.

[Brief description of drawings]

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

FIG. 2 is a diagram showing a modification of a partial pattern / character.

FIG. 3 is a diagram showing a modified example of a character having a partial pattern “mouth”.

FIG. 4 is a flowchart showing the operation of FIG.

5 is a diagram showing a configuration example of a character dictionary used in the device of FIG.

FIG. 6 is a diagram showing a configuration example of a partial pattern dictionary of the apparatus used in FIG.

FIG. 7 is an explanatory diagram of the preprocessing of FIG.

FIG. 8 is a diagram illustrating a pen speed calculation method in FIG.

FIG. 9 is a diagram illustrating a character width calculation from the writing data string in FIG.

10 is a functional block diagram of a pen speed detection unit 20 in FIG.

11 is a flowchart of a pen speed average value v ^* _AVE calculation process in FIG.

FIG. 12 is a diagram illustrating a character width calculation from a feature point data string in FIG.

13 is a functional block diagram of an average stroke length detection unit 30 in FIG.

14 is a flowchart showing the contents of calculation of the average stroke length l _AVE in FIG.

FIG. 15 is an explanatory diagram of a stroke coding process in FIG.

16 is an explanatory diagram of vectors between partial patterns of FIG. 4. FIG.

[Explanation of symbols]

1 tablet 2 pre-processing unit 3 feature point extraction unit 4 stroke coding unit 5 large classification unit 6 medium classification unit 7 partial pattern Q value matching unit 8 partial pattern stroke code distribution matching unit 9 stroke number corresponding parameter setting unit 20 pen speed detection unit 21, 31 x-coordinate interval calculation unit 22, 32 y-coordinate interval calculation unit 23, 33 cumulative addition unit 24, 34 count control unit 25, 35 character width calculation unit 26 cumulative addition / division unit 30 average stroke length detection unit 36 division unit

 ─────────────────────────────────────────────────── ─── Continuation of the front page (72) Inventor Yoko Ikeuchi 1-7-12 Toranomon, Minato-ku, Tokyo Oki Electric Industry Co., Ltd.

Claims (8)

[Claims] 1. A preprocessing unit that removes unnecessary data in a coordinate data string obtained by handwriting on a tablet and performs a linearization process, and a written character from a coordinate data string linearized by the preprocessing unit. A feature point extraction unit that extracts a feature point that represents a feature of a stroke, and a stroke encoding unit that encodes each stroke according to the positional relationship of the feature points extracted by the feature point extraction unit, The output data of the feature point extraction unit or stroke coding unit,
In an online character recognition device that performs character recognition by comparing with registered pattern data registered in advance, a pen speed detection unit that detects a pen speed based on a data string obtained by removing unnecessary data from the coordinate data string from the tablet. An online character recognition device, wherein a threshold value and a weighting parameter in the recognition process are set in advance for each number of strokes, and the threshold value and the weighting parameter are corrected by the pen speed information from the pen speed detecting unit. 2. The pen speed detecting unit according to claim 1, wherein the degree of the continuous character and the degree of the continuous character for each character is stored in advance as a dictionary, and the character which does not match the output from the pen speed detecting unit is stored. An online character recognition device characterized in that some or all of them are deleted from candidates. 3. An online character recognition device comprising a preprocessing unit, a feature point extraction unit, and a stroke coding unit according to claim 1, wherein an average stroke length is extracted from a feature point data string output from the feature point extraction unit. An average stroke length detection unit is provided, a threshold value and a weighting parameter in the recognition process are set in advance for each number of strokes, and the threshold value and the weighting parameter are corrected based on the average stroke length information from the average stroke length detection unit. Online character recognition device. 4. An online character recognition device comprising a preprocessing unit, a feature point extraction unit, and a stroke coding unit according to claim 1, wherein an average stroke length is extracted from a feature point data string output from the feature point extraction unit. An average stroke length extraction / detection unit is provided, and the degree of continuous characters and the degree of continuous characters for each character is stored in advance as a dictionary, and some or all of the characters that do not match the output from the average stroke length detection unit are candidates. An online character recognition device characterized by being deleted from the. 5. An online character recognition device comprising the preprocessing unit, the feature point extraction unit, and the stroke coding unit according to claim 1, after removing unnecessary data of a coordinate data string obtained by handwriting on the tablet. The pen speed is calculated from the data string of, and for each character writing, the calculation result is multiplied by the threshold value and the weighting parameter in the recognition process after the next step which is set in advance for each number of strokes, or is divided by the pen speed calculation result. An online character recognition device characterized in that the character recognition processing is performed after the calculation processing by the above is performed and the threshold value and the weighting parameter are corrected. 6. An online character recognition device comprising a preprocessing unit, a feature point extraction unit, and a stroke coding unit according to claim 1, after removing unnecessary data of a coordinate data string obtained by handwriting on the tablet. The pen speed is calculated from the data string of, and the speed judgment process is performed discretely for each character writing with a certain threshold, and the speed judgment is performed by referring to the continuous character degree stored for each character as a dictionary. An online character recognition device characterized by performing a comparison process with a result after processing, deleting some or all of characters that do not match from candidates, and executing subsequent recognition steps. 7. An online character recognition device comprising a preprocessing unit, a feature point extraction unit, and a stroke coding unit according to claim 1, wherein strokes forming a character are composed of a feature point data string output from the feature point extraction unit. By calculating the average stroke length and normalizing it with the character width, and multiplying it by the threshold and weighting parameters in the recognition process after the next step, which has been set for each number of strokes, or by dividing by the average stroke length calculation result. An online character recognition device characterized in that the character recognition process is performed after performing a calculation process and correcting a threshold value and a weighting parameter. 8. An online character recognition device comprising a preprocessing unit, a feature point extraction unit, and a stroke coding unit according to claim 1, wherein a stroke forming a character is composed of a feature point data string output from the feature point extraction unit. The average stroke length is calculated, and the calculation result normalized by the character width is used to perform the determination process of the average stroke length discretely by a certain threshold for each character writing, and the continuous character degree stored for each character as a dictionary is calculated. With reference to the result after the average stroke length determination process, a comparison process is performed, and a subsequent recognition step is performed after performing a deletion process of deleting a part or all of characters that do not match from candidates. Online character recognition device