JPH01291395A - Alphabetical shorthand recognizing device - Google Patents

Abstract

PURPOSE:To always obtain a sufficient recognition result with respect to a handwritten alphabetical stenograph as an object by detecting the delimitation of each character and the delimitation of each sentence clause, based on a locus of a writing point. CONSTITUTION:A shorthand recognizing means SHRG detects a delimitation between each character of data by an alphabetical stenograph which has been inputted by handwriting, based on at least one of the same point regression of a vector variation speed and a locus appearing in a locus of a writing point at the time of input picture drawing, and a dissipation of the locus. Subsequently, a separation detecting means PNUP detects a delimitation between each sentence clause of data, based on a distance between a dissipation point and a reappearance point of the locus of the writing point. Accordingly, when inputted data can be divided at every character, each character of European language can be recognized by a prescribed pattern matching, and when said data can be divided into each sentence clause, a sentence clause conversion can be executed by taking this opportunity, and each character which has been recognized can be converted to European writing. In such a way, a handwritten stenograph converter which can decide automatically a conversion timing can be obtained.

Description

DETAILED DESCRIPTION OF THE INVENTION [Field of Industrial Application] The present invention relates to a shorthand character recognition device using a computer, and particularly to a shorthand character recognition device suitable for recognizing handwritten European shorthand characters.

[Conventional technology]

Various types of character recognition devices have been known in the past, which use a data processing device such as a personal computer to recognize handwritten data input from a tablet device, etc. , Special Publication No. 57-6151, Special Publication No. 163.8 of 1982
No. 2, Japanese Patent Publication No. 57-19471, or Japanese Patent Application Publication No. 53-45935, Japanese Patent Application Publication No. 55-14095,
This can be found in Japanese Patent Application Laid-Open No. 55-61884.

Incidentally, all of these methods are intended for characters input using the normal writing method, in which each character is written separately.

[Problem to be solved by the invention]

None of the above-mentioned conventional technologies pay particular attention to the recognition of handwritten European shorthand characters, and therefore, when handwritten European shorthand character data is input, the detection of the delimitation of each character and the delimitation of each clause is difficult. There was a problem that character recognition was difficult.

An object of the present invention is to provide a handwritten European shorthand character recognition device that can always obtain sufficient recognition results for handwritten European shorthand characters.

[Means to solve the problem]

The above purpose is to change the separation between each character of data in European shorthand characters input by hand by at least one of the hector change speed appearing in the trajectory of the writing point during input drawing, the same point regression of the trajectory, and the disappearance of the trajectory. This is achieved by detecting the division between each clause of the data based on the distance between the vanishing point and the reproduction point of the trajectory of the writing point.

[For production]

If the input data can be divided into each character, each character in Roman letters can be recognized through a predetermined pattern matching process, and if the input data can be divided into each phrase, this can be used as an opportunity to convert phrases. It is possible to convert each recognized character to Roman characters.

〔Example〕

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS The European shorthand character recognition device according to the present invention will be described in detail below with reference to illustrated embodiments.

Although the embodiments described below are intended for English shorthand characters, the present invention is not limited to English, and can be implemented for handwritten shorthand characters in German, French, and other languages.

FIG. 2 shows an example of a data processing device to which an embodiment of the present invention is applied. In the figure, 1 is a computer device, and 2 is a coordinate manual device.

As is clear from FIG. 2, in this embodiment, a handwritten shorthand character input device is installed in a relatively small computer device, a computer device 1 of the type called a personal computer device for boat fishing, or a business computer device, etc. The system is configured by connecting a coordinate input device 2 which works as a coordinate input device 2. In this embodiment, the coordinate input device 2 includes a tablet and an input pen for drawing characters on the drawing surface of the tablet, as will be described later. Further, the computer system W1 includes the central processing unit 1
a, a memory lb, a display device 1c, a keyboard 1d, etc.

As shown in FIG. 3, the coordinate human power device 2 includes a tablet 2a, a human pen 2b for drawing characters on the drawing surface of the tablet 2a, and a tablet 2a.
A indicates a character input area 2d and an area 2e for specifying various modes on the drawing surface 2C.

2f, 2g, etc. are provided. Then, this coordinate input device 2 inputs data to the drawing surface 2C of the tablet 2a, and inputs data to the pen 2b.
The function is to output status information indicating whether or not the tip of the writing point is in contact with the writing point, and X coordinate value information and Y coordinate value information indicating the contact position of the writing point, as necessary. Note that the coordinate input device 2 is not limited to a tablet, and any coordinate input device that can send out the above output information may be used.

As characters are written in the character input area 2d with the input pen 2b, X and Y coordinate value information as data regarding the writing point is sequentially output. Various mode settings are instructed by pressing the input pen 2b against the corresponding areas 2e, 2f, 2g. In memory 1b,
A shorthand character recognition means, a distance detection means, a shorthand English phrase conversion processing means, and various programs for effectively starting each of these processes are stored, and by processing and executing the programs, the central processing unit 1a Accomplish a given function. Furthermore, this memory 1b has various types of storage for storing various data from the coordinate input device 2 necessary for the central processing unit 1a to process each of these programs, and various intermediate data necessary for the processing. act to provide the department. In this embodiment, the display device 1c is used to sequentially display handwritten handwriting or to display recognized characters.

FIG. 1 shows the principle of the present invention, and is a block diagram showing the mutual relationship between the above-mentioned constituent parts. A is a handwritten shorthand manual means;
This corresponds to the coordinate input device 2 shown in the figure. 5HRG is shorthand character recognition means, PNUP is separation detection means, and EJC
N is a shorthand English phrase conversion processing means. These means 5H
R.G.

As described above, PNUP and EJCN are stored in the memory 1b in the form of programs, and when the central processing unit 1a processes and executes the programs, they perform predetermined functions.

In FIG. 1, the handwritten shorthand character input means A sequentially inputs data regarding the tip contact portion of the input pen 2b, that is, the trajectory of the writing point, to any part of the drawing surface 2d on the tablet 1-28 of the coordinate input device 2. Output.

The shorthand character recognition means 5HRG is the handwritten shorthand character input means A.
The output data from the computer is input sequentially, and handwritten shorthand characters are sequentially recognized based on the data.

Similarly, the distance detection means PNUP sequentially inputs the output data from the handwritten shorthand character input means A, and based on the data, the tip of the input pen 2b, that is, the writing point, is on the drawing surface 2d.
Detect whether or not the object is in contact with the object. Note that the presence or absence of this contact is called separation.

The shorthand English phrase conversion processing means EJCN manually inputs the character data recognized and verified by the shorthand character recognition means 5HRG, and responds to the pen point separation distance detection information from the distance detection means PN [JP] based on the character data. Executes English clause conversion processing.

Thereby, it is possible to obtain a handwritten shorthand character conversion device that can automatically determine the conversion timing without requiring the operation of a conversion instruction key or the like.

Next, the operation of each of these means will be explained in detail below.

The details of each processing means will first be explained in detail with reference to FIG. FIG. 17 is a memory configuration diagram showing the configuration of a primary storage section necessary for execution of each processing means, and each storage section is set at a predetermined address in the memory 1b.

GDST is a coordinate value information acquisition status storage unit that stores the acquisition status of each signal received from the coordinate input device 2, that is, a status signal, X coordinate value information, and Y coordinate value information.

STT, XDT, and YDT are storage units that respectively store the status information, X coordinate value information, and Y coordinate value information received from the coordinate input device 2.

PNP is a previous pen state storage unit that stores status information associated with the coordinate values of the writing point of the input pen 2b in the past, that is, whether it is in the pen completed/down state or in the pen down state;
ST determines the state of the input pen 2b, that is, the pen-down time, from the contents of the current status information storage section ST''F and the contents of the closest past status information storage section PNP.
Pen down.

This is a pen state storage unit that stores the state of the pen at the time of pen-up.

The DOTS is a coordinate system that sequentially stores status information, X coordinate value information, Y coordinate value information, and various information obtained from these received from the coordinate input device 2, and sequentially and cyclically stores data regarding the trajectory of a writing point. This is a value storage section. This coordinate value storage unit DOTS is logically configured in a ring shape and stores information regarding the locus of the writing point in a circular manner.

The STCM is a character code storage unit that sequentially stores and accumulates character codes of stored shorthand characters.

Below, with reference to this Figure 17, details for each figure -8= Explain.

4 and 5 are flowcharts showing the main processing means, and FIG. 5 is a trajectory acquisition means for inputting various information output from the coordinate input means 2 to the computer device 1 using an interrupt function. In this process, first in step 5a, as will be described in detail later, the coordinate acquisition means GD is activated and the coordinate information of the writing point at the relevant time point and the status information of the input ben 2b are acquired. , are sequentially stored in the coordinate value storage unit DOTS. Next step 5
In step b, as will be described in detail later, the pen state acquisition means PS is activated, the state of the input pen 2b is determined from the stored contents in step 5a, and this is similarly stored in the coordinate value storage section DOTS.

In the main processing means shown in FIG. 4, first, in step 4a, initialization means INIT is activated. This means INIT
As will be described in detail later, various time storage units and each part of the device are initialized. In the subsequent step 4b, the character delimiter detection means 5PDT is activated, as will be described in detail later. This means SPDT detects the delimitation of each shorthand character based on the data taken in by the locus acquisition means TG shown in FIG. That is, as shown in FIG. 18, in shorthand characters, multiple characters are written in one stroke, so when recognizing each character, it is necessary to detect the break in order to separate each shorthand character.

In this manner, the character delimiter detection means 5PDT is executed in step 4b, and when a delimiter of shorthand characters is detected, the shorthand character recognition processing means 5HCR is activated in step 4C. This means 5HCR responds to character delimitation detection information from the character delimitation detection means 5PDT and recognizes shorthand characters based on the data of shorthand characters minus characters.

Then, in step 4d, the character codes of the recognized characters are sequentially stored in the character code storage section STCM.

In the following step 4e, the input pen 2b is a tablet) 2a
It is determined whether or not the character has left the drawing surface 2c. If the character has not left the drawing surface 2c, the process returns to step 4b and recognition of the next character is executed. However, after pen-up is detected,
When the next pen-down is detected and the series of operations in which - character recognition is completed, the process proceeds to step 4f, where the pen-up period distance is determined, and if the calculated distance is less than a certain value, the pen-up It is assumed that writing of a passage including motion is continuing, and the process returns to step 4b to start character recognition again.

If the pen-up distance is equal to or greater than a certain value, the process proceeds to step 4g, and English clause conversion EJCN is performed based on the character code before pen-up stored in the character code storage unit STCM.
Execute. Then, the converted phrase is passed to a program such as a document creation device, and in step 4h, it is determined whether or not to end the series of processing. If not, the process returns to step 4b and the series of processing is repeated. repeat.

By executing such a series of processes, handwritten shorthand characters are converted into English text.

Note that in correspondence with FIG. 1, step 4b.

Steps 4c and 4d correspond to shorthand character recognition means SHRG, and steps 4e, 4f and 4j constitute separation detection means PNUP.

Next, each processing means in FIGS. 4 and 5 will be explained in detail.

FIG. 6 is a flowchart showing details of the coordinate value acquisition means CD outlined in step 5a of FIG.

Upon activation, the means CD first receives various information from the coordinate input device 2 in step 6a, and in the subsequent step 6b, identifies which information the means CD is waiting for input in its internal state. . Here, the storage contents of the coordinate value acquisition state storage unit GDST are: when it is "0", it is waiting for status information; when it is "1", it is waiting for X coordinate value information;
When it is "2", it is predetermined to wait for Y coordinate information.

Next, in steps 6c, 6f, and 6j, the data types of input information and waiting information are checked. In particular, in steps 6f and 6j, when input data of different types is detected, the internal state is reset by clearing the storage section GDST in steps 6i and 6m.

Furthermore, in steps 5d, 6g, and 6, the received information is stored in the corresponding storage units STT and XDT, respectively.
, transferred to and stored in YDT, step 6e, step 6h,
In step 61, the storage unit GDS
Rewrite T's memory contents.

Finally, in step 6n, the storage unit STT
, XDT, and YDT are transferred and stored in the coordinate value storage section DOTS.

Therefore, this coordinate value acquisition means GD sequentially imports information from the coordinate manual means 2 in the order of status information, X coordinate value information, and X coordinate value information, and stores each retrieved information in the corresponding storage units STT, XDT, and YDT. Having received and stored all of this information, it serves to prompt the execution of subsequent processing means.

Next, FIG. 7 is a flowchart showing the details of the pen state acquisition means ps, which was outlined in step 5b of FIG.

Upon activation, the means PS first refers to the stored contents of the status information storage section STT in step 7a and determines whether the input pen 2b is currently in the down state.

Next, in steps 7b and 7c, a change in the state of the input pen 2b is detected. That is, the current pen state, that is, the stored contents of the status information storage section STT and the previous status information storage section PNP are compared, and if they are in the pen-down state and pen-down state, the pen-down is continuing, and the pen-down state and the pen-up state are respectively determined. If so, it is determined that the pen is down, and if there is a pen up state and a pen down state, it is determined that the pen is up.

Next, in step 7d, step 7e, and step 7f,
Information indicating the corresponding states is stored in the pen state storage unit PS.
Transfer and store in T. Furthermore, if there is a state change in the pen state, i.e. from pen up state to pen down state,
Alternatively, when the pen-down state changes to the pen-up state, in corresponding steps 7g and 7h, information indicating the current pen state is transferred and stored in the previous pen state storage unit PNP in preparation for the next process.

Finally, in step 71, the contents of the storage section PST are stored in the coordinate storage section DOTS in the same manner as described above.

By executing this pen state acquisition means PS, it is possible to identify whether the current pen state is at the pen down time, pen down continuation, or pen up time.

As described above, the coordinates of the trajectory of the writing point and the state of the manual pen 2b are sequentially stored in the storage unit DOTS by the coordinate acquisition means GD and the pen state acquisition means ps shown in FIGS. 6 and 7.

FIG. 8 is a flowchart showing details of the initialization means INIT, the outline of which was shown in step 4a of FIG.

As described above, the means INIT functions to set the device and the time storage unit to the initial state, but in this embodiment, the coordinate input device 2 is connected to the computer device 1 via a communication port. Therefore, in step 8a, first set the baud rate of this communication boat to, for example, 9.
600 CBPS). Normally, the baud rate of the keyboard 1b is 9600 CBP SE, so to maintain compatibility with this, the baud rate is set to the same.

The next step 8b is to disable the interrupt by keyboard 1" 1b, so that instead of keyboard 1b,
Enables input of data using the coordinate input device 2. Then, the recognized character is inserted into the system buffer managed by the system.

In the following step 8c, storage units as various work areas are initialized, and in step 8d, a memory buffer for storing the recognized shorthand character code is initialized. In this example, 30 characters are reserved.

Next, in step 8e, the position of the cursor on the display screen of the display device 1G is stored.

Finally, in step 8f, the participants are asked to decide the switch operation for general character recognition. The switch operation method is such that shorthand characters are not misrecognized, for example, by pressing the protrusion on the side of the pen, or by pressing the dot two or more times in succession. Alternatively, there are other methods such as changing the pressure of the pen.

For this reason, in step 8e, the switching method is displayed on the screen, and the user is asked to specify the type of switch operation from among the switching methods.

Next, the shorthand character recognition process will be described.

The basic character translation of shorthand, the kanji, includes the angle and hector change between dot rows of shorthand characters, recognition of the same point, recognition by distinguishing the length of shorthand characters, the dora 1 of one character of shorthand characters - n of the row
Recognition based on the absolute angular difference and relative angular difference of point-approximated points, and recognition using the correlation method and distance method for the n-undivided dots of shorthand characters are known, but in each case, The most suitable recognition method will be adopted accordingly.

FIG. 9 is a flowchart showing details of the character delimiter detection means 5PDT shown in step 4b of FIG. In this embodiment, character breaks are detected using the following two methods.

■ Character break detection method (i) Recognition based on rapid changes in the vector between shorthand points Trajectory means T
G is sequentially stored in the coordinate value storage unit DOTS. Therefore, in step 9a, this coordinate value information is sequentially read out, and the coordinate value of the position of the writing point (Xi + Y,) and the coordinate value of the next writing point (xt. +, y = -) are obtained.
+) is calculated using the following equation (1) and stored. Hereinafter, this θ will be referred to as a hector between shorthand points, or simply a vector for short.

In the following step 9b, the vector value θm detected in step 9a and the previously detected vector value θ, -5
The difference between the vector θ1 and the vector θ1 as shown in FIG. 18 is thereby detected.

In this embodiment, the sudden change value of the vector θ1 at this time is set to 60° or more.

In this way, when a point SP is detected where -·ctor θ is changed by 60° or more, this point SP is determined as the delimiter of one character. In this embodiment, the character delimiter detection flag MFI
is set, and when point SP is detected, this flag M
Set FI to rlJ. Note that this flag MFI is used when character recognition is performed based on the angle of the character, which will be described later.

(11) For the shorthand point-to-point vector θ, −0 when the shorthand point-to-point vector and the trajectory of the recognized writing point by same-point recognition pass through the same coordinate point, and the first trajectory passes through, The shorthand point-to-point vector θ when the trajectory passes next,
indicates a difference of a predetermined angle W or more, it is determined that a character is separated.

1θ, -θ, -11≧W In this example, it is set to W-60°.

Also, at this time, -180 6<θi<180.

-1806<θi, I<180.

shall be satisfied.

The above judgments are processed in steps 9c and 9d. At this time, in English shorthand, it is necessary to separate the information before the vowel, so the process of step 9f is performed.

■ Recognition by pen-up When a pen-up is detected, it is immediately assumed that a break has occurred.

This processing is then provided in step 9e.

Character breaks are recognized using the above three character break detection methods, but if a character break is not detected, the process returns to step 9a and the angle between the writing point on the clothing and the pen up is again determined. Performs single point detection.

When a character break is detected in this way, the shorthand point-to-point vector θ
, and the locus coordinates of the writing point xi, )'i can be used as one character to separate data. Note that the start of writing can be detected by pen-down information, and the end of writing can be detected by pen-up information.

FIG. 10 is a flowchart showing details of the shorthand character recognition processing means 5HCR, the outline of which was shown in step 4c of FIG.

First, we will discuss the recognition of the length of shorthand characters.

Examples of shorthand characters are shown in Figures 22 and 23. As can be seen from these figures, in shorthand, long lines are generally written approximately three times as long as short lines.

Therefore, in order to distinguish between long lines and short lines, in this example, short-line shorthand characters and long-line shorthand characters are usually distinguished based on 1.2 [(to)]. There is.

However, there are exceptions, such as the F and V of the A concave in Fig.
Examples include CH and J in XA-, P and B in E-concave, etc.

Therefore, in this case, the boundary values are set at twice and three times the shorter character. That is, in general, 1. (c
This is because characters that are longer than m) have an average length of 2 [Cm], and characters that are shorter than that have an average length of 7 (1 m).

However, since the font size differs depending on the stenographer, this embodiment is provided with a learning mode in which the scribe can specify the boundary value for distinguishing long and short characters, and step 10e
Now, in order to select this learning mode, determine the switch that is a separate membrane, and if the switch is on, proceed to step 10.
Start the learning means STUM with f. FIG. 12 shows a detailed flowchart of this learning means STUM.

First, in step 12a, a character in which the sum of relative angles and the sum of differences in angles are similar is specified, and the scribe is requested to write the character in the size that the scribe would normally write. Then, in step 12b, the written characters are displayed as they are on the display device 1c, and the scribe himself/herself is made to recognize the size of each character using the same shorthand character angle.

For example, if the scribe writes letters that differ only in size, such as the letters "N" and "M", this is determined in step 12d. Then, in step 12c, a message indicating the error is displayed on the display device, and a re-input is requested. By repeating these steps, in step 12e, the point at which the length of the scribe's handwritten characters can be distinguished is recognized and memorized. Note that the above-mentioned "character size" is determined by the following method in this embodiment.

First, in step 10c, the x and y coordinates divided into one character are equally divided and approximated into N sample points. In addition, in Example 2, it is N-12.

Figure 21 is an example of rVJ shorthand characters, which are 12
The figure shows the case where the approximation is made by equally dividing the sample points.

Next, as shown in FIG. 25, the barycenter coordinates (X, Y) are determined from the N sample points recognized as one character using the following equation.

From the center of gravity thus determined, the distance to the sample point with the greatest distance is determined using equation (3), and the length β is taken as the size of the character.

The above recognition process is step 10g.

Next, we will explain the principle of angle recognition.

The angle at which the shorthand characters are written is a circle as shown in FIG. 19 or a circle as shown in FIG. 23. And the shorthand characters are
Basically, shorthand characters are drawn at the angles described above, so in this embodiment, the following two recognition methods are used to recognize shorthand characters. The angular character recognition means AGRG shown in step 10h of FIG. 10 executes this.

-24= At this time, a switch for general character recognition is provided in advance and is turned on, so it is removed in step 10d and step 107! is moving to.

FIG. 11 is a detailed flowchart of the angular character recognition means AGRG, which performs recognition using two methods as described above.

(2) Recognition based on the sum of relative angles First, in step lla, the relative angles of the sample points are calculated. In FIG. 20, the angle θ is the starting point Ps
(xs, y5) to sample point P, (i=1 to 6)
Indicates the angle to. This angle θ is calculated by the following equation (4).

Next, the state of the MFI mentioned above is checked, and if the MFI is not 1, the process moves to the next step.

When MFI is 1, the following processing is performed.

Next, character information is separated by salient point recognition.

As shown in Figure 26, the above-mentioned process 5P of recognizing separation of one character even though multiple characters are written.
When this is not detected in DT, check the increase/decrease direction of the value obtained by equation (4), and separate the points where the increase/decrease direction is reversed or suddenly increases/decreases as one character. recognized as

At this time, the process returns to step 10c in FIG. 10 in order to separate the XY coordinate data character by character.

In step Llc, the sum of relative angles is calculated using the following equation (5) from the values obtained using equation (4).

Next, in step 11d, the angle type of the shorthand character and the unevenness of the character are detected based on these calculated values.

FIG. 22 is a diagram in which the shorthand characters shown in FIG. 19 are classified according to length, type of angle, and unevenness, and the shorthand characters are recognized according to such classification. Furthermore, in this figure, the corresponding shorthand character ``Mion'' is added in parentheses. In the subsequent step lie, the sum of the differences in the angles of the sample points is calculated.

■ Recognition based on the sum of angle differences, that is, the angle θi (i = 1 to 10) in Fig. 21 is the difference between the vectors connecting each sample point, and in step 11e, the sum of the angles θ1 is calculated. That's what I do.

The sum of relative angles and the sum of differences in angles calculated as described above are compared with corresponding angle values in a standard dictionary created in advance in step ILf, and characters with a small angular difference between the two are selected in step l1g. Select and memorize.

Returning to FIG. 10, in step 10g here,
The size of the shorthand character is recognized as described above, and in the following step 10h, the shorthand character is recognized based on the size of the character, the sum of the above-mentioned relative angles, and the sum of the above-mentioned difference degree difference of the angles, and a plurality of shorthand characters are recognized. Select candidate characters. In step 10i,
The character recognition means BRGM using the approximate direction vector series method is activated.

The means BRGM, as shown in detail in FIG. 13, correctly recognizes shorthand characters from a plurality of selected candidate characters by an approximate directional vector series method of the following segments. As pre-processing, the means BRGM converts the sample point example data into data of the size of a standard dictionary of candidate characters in step 13a. and,
The following calculation based on the approximate direction vector sequence method is performed.

The approximate direction vector is the difference value itself between both end points of the segment. -If the (N+1) feature points obtained by dividing the character into N equal parts are p, (starting point) to P N+1 (end point), then
The approximate direction vector ΔP8 is ΔPi=P, +, -p.

= (Xi+, -X8.Y, +IYt) - (6 squares, i-1~N, P, l= (XIl, Yfi) n=
l to N+1.

5/1 is the stroke between human characters and the standard pattern, and the distance ds is
It is calculated as the sum of distances between feature parameters.

That is, ds=llP1-P, 11... Distance between starting points +
1PH-I PN+l II... Distance between end points.

In step 13b, the standard dictionary pattern of candidate characters obtained according to the size and angle of the characters and the scribe's shorthand characters are calculated using the above equation (8), and the character with the minimum parameter distance is selected.

In step 13c, if the distance between the approximate direction vectors obtained from the above equation (7) between the selected character and the scribe's handwritten character is an approximate value, it is stored that the character has been recognized in step 13d. , otherwise it is stored in step 13e that the character is not recognized.

Then, returning to FIG. 10 again, step IQk is executed. This step 10 is a process of determining whether or not a character has been recognized. Specifically, step 13 in FIG.
d or the contents stored in step 13e. Here, if the characters have been recognized, the execution of the shorthand character recognition 11isHRG shown in FIG. 10 ends. On the other hand, if no characters are recognized here, step 1
01 starts the special character recognition means 5CRG. FIG. 14 is a flowchart showing details of this special character recognition means 5CRG.

The means 5CRG in FIG. 14 is shown in FIG.
As shown in the figure, special types of characters are determined, such as when the pen is separated during character writing, and character recognition is performed accordingly.

First, as determined in step 14a, if this special character recognition is called for general character recognition, step 14I of general character recognition! Move to. Otherwise, the process moves to step 14d.

Here, the angular character recognition means in step 14d, the character size recognition means in step 14e, the means for selecting a recognition character candidate group in step 14f, and the character recognition processing using the approximate direction hector sequence method in step 14g will be compared. By simply replacing the dictionary data with special character data instead of shorthand character data, step 10h of 5HCR in FIG. 10, Log.

Perform the same operation as in step 10j.

In the next step 14h, character group conversion recognition processing is performed for the characters recognized in the previous steps. FIG. 16 shows a detailed flowchart of this character group conversion recognition process 5CHRG.

First, when it is determined in step 16a that it is not a point,
In the next step 16b, the character just recognized is the 27th character.
When the vowel [i] is recognized as shown in part (a) of the figure, the recognized character code of the previous character, which has already been recognized and stored in the system buffer, is replaced with the corresponding character code.

On the other hand, in the next step 16c, character combinations not recognized by 5CRG are recognized. For example, in the case of the character "?" as shown in Figure 28, the character delimiter is recognized at the sp part in the figure, and all of the characters are combined and recognized as one character as shown in Figure 29, so step In 16c,
Match the pattern group of the special character that is already a combination with the pattern group recognized so far, and if it is still not recognized, the user's special character is tW, th if it is a special character with a stroke count of 3, and is recognized in the step of 5CRG. The resulting pattern code is stored and the process returns to the step of FIG.

Then, in step 14j, the number of strokes of the drawn character is checked, and if the number of strokes is greater than the maximum number of pixels listed in the special character dictionary, an error is displayed. Before that, the switch is checked in step 14i, and if the general character recognition processing switch is on, the process advances to step li.

Now, returning to FIG. 4, after storing the character code in step 4d, the pen-up counter is checked in step 4j, and if it is on, the process moves to step 4f to detect the pen-up period distance, and if it is off, the pen-up counter is checked. Proceed to step 4e.

In step 4e, pen-up detection is performed, and if the pen is not up, the process returns to step 4b and character detection is performed again.

If pen-up is detected, the process advances to step 4f.

Step 4f is a pen-up period distance detection process PLD, the details of which will be explained using the flowchart of FIG. 15.

In step 15a, the pen 7 counter is counted eight times, and if it is not on, the pen up counter is turned on in step 15c, and in step 15e, the pen up distance is made small, and the process moves to step 4b in FIG.

If the pen-up counter is on, in step 15b, step Lo of the character group recognized before pen-up is set.
Find the average value of the center of gravity found in g. Next step 15b
The distance between the value obtained in step 15 and the center of gravity of the character obtained after pen-up is obtained in step 15d, and if the value is greater than a certain value, the pen-up period distance is set to 1 large.
Move to 5f and turn off the pen up counter.

As described above, the magnitude of the pen-up distance is determined, and if it is small, the process returns to step 4b and character recognition is started again.

Therefore, according to this embodiment, one phrase is handwritten on the drawing surface 2d of the tablet 2a, and upon completion of the writing, the input pen 2b
is removed from the drawing surface 2d and character separation and character recognition are performed again for one character. Then, the distance between the center of gravity of all the phrases before pen-up and the center of gravity of one character after pen-down (pen-up period distance) is calculated. If the value is above a certain value, transfer the one phrase before pen-up to the English speech compatible processing means in step 4g,
One phrase recognition is performed, and if the value is less than a certain value, the process returns to step 4b, and the operation of recognizing shorthand characters is subsequently obtained.

In other words, by detecting the pen-up period distance, the series of operations of inputting each phrase and turning on the "recognition" switch is no longer necessary, and the user can perform speed entry without being aware of the tablet. I can do it.

Next, the shorthand English phrase conversion processing means EJCN (step 4g) in FIG. 4 will be explained.

This step 4g is an English phrase processing conversion that recognizes the shorthand character codes that have been stored as phrase codes as phrases, converts them into accurate characters, and outputs them. Figure 30 shows the detailed flowchart. .

First, in step 30a, the number of characters in a clause is calculated.

Then, if there is a pen-up in the middle of the phrase as determined in step 30b, steps 30c and 30d are executed,
The maximum number of characters searched for a keyword is set to the point at which pen-up is recognized.

In step 30e, the keywords are compared with keywords (character combination patterns used when omitting) in the dictionary shown in FIG. 31 in order from the front and the back.

When a keyword is found, the process moves to step 30g and a search for characters is performed using the keyword dictionary (FIG. 31).

In step 30h, increase the number of search characters in the forward and backward directions,
Then look for keywords that have one more character than before. When a series of operations is searched up to the maximum number of keywords and the corresponding keyword is not found, or when there is no corresponding phrase in the keyword dictionary, the search is performed from the beginning in a legitimate dictionary in step 30j, and when the keyword is found, , in step 30, it is converted into a valid phrase in English, and if there is no corresponding character, in step 4d, the stored character code is returned as is without conversion.

Therefore, according to this embodiment, when entering shorthand characters,
The above-described process is repeated, and the shorthand characters are recognized and converted into English, allowing the tablet to recognize English shorthand characters.

Although the above embodiment describes an example in which the present invention is applied to the recognition of English shorthand characters, Kitoaki is not limited to English shorthand characters, but can also be applied to shorthand characters of other European languages. It goes without saying that this may be implemented.

〔Effect of the invention〕

According to the present invention, since it is possible to recognize the position of one character in European shorthand characters, it is possible to immediately judge the characters that have been manually written. In addition, the timing of Roman phrase conversion can be automatically determined by detecting the distance between the pens, so shorthand Roman phrase conversion can be performed without the need to operate a conversion instruction key, etc., and furthermore, shorthand can be omitted. Searching for shape patterns enables faster matching with dictionaries,
Shorthand characters can be recognized more easily and quickly.

[Brief Description of the Drawings] Fig. 1 is a functional block diagram showing an overview of the present invention, Fig. 2 is an explanatory diagram of a character recognition device to which an embodiment of the present invention is applied,
FIG. 3 is an explanatory diagram showing an example of a coordinate input device, FIGS. 4 and 5 are flowcharts showing the main processing means in an embodiment of the present invention, and FIG. 6 is a flowchart 1/1 showing coordinate acquisition processing GD. , FIG. 7 is a flowchart showing the pen state acquisition processing ps, FIG. 8 is a flowchart showing the initialization processing INIT, FIG. 9 is a flowchart showing the character break detection processing 5PDT, and FIG. 1Q is a flowchart showing the shorthand character recognition processing 5HCR. The flowchart shown in FIG. 11 is angular character recognition processing AG.
Flowchart showing RG, Figure 12 is learning processing STU
FIG. 13 is a flowchart showing character recognition processing BRGM using the approximate direction vector series method, FIG. 14 is a flowchart showing special character recognition processing 5CRG, and FIG. 15 is pen-up separation distance detection processing P.
LD flowchart, Fig. 16 is a flowchart 1 showing the character delimitation detection process S CHRG, Fig. 17 is a configuration diagram of the memory used in various processes, Fig. 18 is an explanatory diagram of character delimitation recognition, Fig. 19 The figure is an explanatory diagram of the composition angle of shorthand characters,
Figures 20 and 21 are explanatory diagrams of angular character recognition, Figure 22
Figure 23 and Figure 24 are explanatory diagrams of English shorthand characters,
Fig. 25 is an explanatory diagram of the principle of character size recognition, Fig. 26 is an explanatory diagram showing an example of two shorthand characters written in one stroke,
27, 28, and 29 are diagrams explaining the principle of special character recognition, number 30 is a flowchart showing the shorthand character recognition process EJCN, and FIG. 31 is a diagram explaining the dictionary used in the EJCN process. 1... Computer device, 2... Coordinate input device, 2
a...Tablets 1~. 2b... input pen, 2C...
・Drawing surface, 2d...Character input area, 2e to 2f...
Mode setting area. σr Fig. 8 Fig. 9 Fig. 10 Fig. 5
Monme row IJ near flea person size 0d general explosive song/call YES No +0fSTυ
M YES to the human nature of literature I have a lot of learning methods O I recognize the thick can of potatoes, *su6 oh GRG Corner & length, literacy °
0, r & death YES 1/'11) Figure 11 TNO5CRG ↓ /゛-゛Van1! L
J'7 Figure 2g (Gu. !! Gu, Water

Claims (1)

[Claims] (1) A means for detecting the positional coordinates of the point contact portion with respect to the detection surface, which sequentially captures data regarding the locus of the writing point given the detection surface as the drawing surface and divides the data for each character of handwritten shorthand characters. In a European shorthand character recognition system that recognizes the data and the data delimitation for each clause, the detection of the data delimitation for each character is performed by using the vector change speed of the trajectory and the same point regression of the trajectory, and A European language characterized in that the detection is performed based on at least one of the disappearances of the trajectory, and the detection of the data break for each clause is performed based on the distance between the vanishing point and the reappearance point of the trajectory. Shorthand character recognition device.