JPH069064B2 - Handwriting recognition device - Google Patents

Description

Description: TECHNICAL FIELD OF THE INVENTION The present invention relates to an online processing handwritten character for recognizing an input character by inputting, for example, a character input handwritten on an input tablet and executing the dictionary collation processing. Regarding the recognition device.

<Background of the Invention> As a conventional character recognition device of this type, for example, as disclosed in Japanese Patent Laid-Open No. 58-200380, in preparation for a case where a character with an input frame for character input runs out. , If the detection area is expanded in advance by a certain width so that the handwriting written in a certain area around the entry frame can also be detected, and if the area where the handwriting is extended also protrudes,
There is a device that extracts the handwriting feature of the boundary portion of the expanded area and uses it as one of the judgment materials for character recognition.

However, in such a device, it is essential that the device be provided with a means for extracting and analyzing the handwriting feature at the boundary in case the handwriting overflows a predetermined area. Had complicated the structure of.

<Object of the Invention> The present invention does not require a means for specially extracting and analyzing the handwriting feature at the boundary even when the handwriting largely extends out of the entry frame, and character recognition can be performed by ordinary dictionary matching. An object of the present invention is to provide a character recognition device having a simple structure that can be performed.

<Structure of the Invention> The handwritten character recognition device of the present invention has an input tablet for detecting whether a pen-shaped object has come into contact with or separated from, by forming a grid for writing one character in an area for inputting characters by handwriting. And a recognition processing unit that executes a character recognition process by taking in handwritten character input to the input tablet, and the recognition processing unit has a predetermined extended area around the grid as a basic area for pen position detection. Setting means for setting, an enlarging means for enlarging an area for cutting out one character stroke based on the handwriting in the basic area and the handwriting continuing from the basic area to the extension area, and the handwriting in the cutout area enlarged by the enlarging means. A storage means for converting the whole coordinate and storing it in the basic area.

<Effects of the Invention> According to the present invention, even with respect to a handwriting that greatly protrudes a square, which is a basic area, the entire handwriting can be coordinate-converted and stored in the basic area, and then character recognition can be performed by normal dictionary matching. It is possible to provide a handwritten character recognition device having a simple structure and high accuracy of dictionary matching as compared with the above-mentioned conventional technique.

Also, since the area where one character stroke is cut out is enlarged on the condition that the handwriting continues from the basic area to the extended area, even if there are stains due to noise in the extended area, this is different from the characters in the basic area. Can be correctly recognized as a thing.

Furthermore, because the coordinates of the entire handwriting in the clipping area are converted,
There is no risk of changing the shape of the entire character, and the recognition rate is not reduced.

<Description of Embodiments> FIG. 1 shows the overall configuration of a handwritten character recognition device.

The apparatus shown in the drawing is a personal computer main body 1 (hereinafter simply referred to as "computer main body"), an input tablet 2,
A printer 3, a floppy disk device 4, etc. are included, and these respective components are electrically connected by a cord line. The computer main body 1 is equipped with a CRT (Cathode Ray Tube) 5 and a keyboard 6, and the keyboard 6 is provided with a plurality of function keys for inputting various commands and the like. The command input section is not limited to the keyboard but may be provided on an input tablet described later and a pen operation method may be adopted. A CPU (Cent
RAM (Random Access Mem)
ory), ROM (Read Only Memory), and other recognition processing units are included, and this recognition processing unit controls the input / output operations of each input / output unit and executes a series of character recognition processing including dictionary collation. To be done.

A character input area 7 is provided on the upper surface of the input tablet 2 for inputting characters by handwriting. A plurality of vertical and horizontal squares 8 are formed in the character input area 7, and each square 8 is formed.
Then, the input pen 9 is brought into contact with and handwritten characters are input.
The input pen 9 has a structure in which, when the pen tip touches the character input area 7 and a predetermined writing pressure is applied, for example, magnetic force lines are generated at the pen tip portion, and accordingly, the magnetic force lines move in the input tablet 2. By detecting the locus, the handwriting from writing to releasing is obtained as character stroke data.

According to the present invention, a grid for pen position detection is used as a basic area, and a predetermined extended area is set around the basic area, and a handwriting protruding to the extended area is stored in the basic area under a certain condition. The area in which one character stroke is cut out is regarded as a handwriting and is enlarged. The constant condition referred to here is a condition caused by the continuity of the handwriting that protrudes. For example, as shown in FIG. 2, when the handwriting a in the basic area A has a continuous protruding part b, the protruding part b is continuous. The stroke cutout surface M is enlarged to a position where b has entered the extended area B most deeply (the right limit c shown by the one-dot chain line in FIG. 2), and all handwritings included in the stroke cutout surface M, that is,
With respect to the protruding portion b which is continuous with the handwriting a in the basic area A and the handwriting d which is discontinuous with respect to the basic area A but is included in the stroke cutout surface M, these are regarded as the handwriting in the basic area and the stroke It is to recognize. In this case, when there are a plurality of handwritings protruding from the basic area A in the same extended area B, the stroke cutout surface M is enlarged to the position where the handwriting penetrates the extended area B most deeply among the handwritings. To On the other hand, the third
As shown in the figure, when the protruding portion b'is written completely separated from the basic area A, the above treatment does not work.

The range and size of the extension area can be set as appropriate.
In the example shown in the figure, eight squares on four sides adjacent to the basic area A (indicated by a thick frame in the figure) are expanded areas B (indicated by diagonal lines in the figure), and in the example shown in FIG. In the example shown in FIG. 6, four squares in the upper and lower sides and the right side are set in the extension area B, and in the example shown in FIG.

Thus, when a character stroke protrudes from the basic area A to a predetermined expansion area B, the handwriting including the protruding portion is drawn in the basic area A by using one of the methods illustrated in FIGS. 7 (1) to (4), for example. After that, one character stroke is cut out, and character recognition processing such as feature extraction and dictionary matching is performed. FIG. 7 (1) is a method of reducing the horizontal length XAB of the basic area A and the extended area B to 1/2, and FIG. 7 (2) shows that the maximum horizontal length x of the character stroke is less than the length XA of the basic area A. Sometimes the character stroke is moved laterally to the left. In Fig. 7 (3), when the maximum horizontal length x of the character stroke is the horizontal length XA of the basic area A or more, the maximum horizontal length x matches the horizontal length XA. 7 (4) is the basic area A
And the horizontal length XAB and vertical length YAB of the expansion area B are 1 /
It is a method of reducing to 2.

FIG. 8 shows the contents of the memory in the recognition processing unit,
A program area 10 in the figure stores a series of programs for character recognition processing. The work area 11 includes a coordinate storage area 12, a stroke code storage area 13, a square number setting area 14, and an extended area information storage area 15.
Etc., the coordinate storage area 12 contains the pen position coordinates of the input pen 9 on the input tablet 2, and the stroke code storage area 13 contains the 8-direction code (shown in FIG. 9) for the character strokes. The setting area 14 stores the square number of the basic area, and the extended area information storage area 15 stores the usage status of the extended area. In addition, the character storage area 16 stores the recognition result of all input characters, and the dictionary area 17 stores standard data for dictionary collation for kana and alphanumeric characters.

FIG. 10 shows an operation processing flow of the recognition processing section. First, in step 20, a stack pointer is initialized to the memory of the recognition processing section, and the contents of the work area 11 are cleared. Next, step 21 is the computer main body 1
It is determined whether or not there is a key input from the keyboard 6 of No. 6, and when the key requesting the tablet input is pressed now, the determination of "command input is present" in Step 21 and "input start" in Step 22. Both of the judgments of
When the input pen 9 comes into contact with the character input area 7 of the input tablet 2 and a predetermined writing pressure is applied to the character input area 7 of the input tablet 2, the determination of "writing" is made in step 23.
S ”, and the number of the square of the pen tip is the data indicating the basic area, and the square number setting area 1 in the work area 11
4 and the coordinates of the pen position in the grid are written in the coordinate storage area 12 (steps 24, 2).
5). Further, when the pen tip moves and a character stroke is drawn in the same square (basic area), the determination of "whether or not to write" in step 26 is "NO", and the determination of "out of the basic area" in step 27 is also " If NO, the process returns to step 25 and the coordinates of the constituent points of the character stroke are stored in the coordinate storage area 1
2 are sequentially written. At this time, if the handwriting extends beyond the basic area to a predetermined extension area, the determination in step 27 is "YES", and the determination "whether it is in the extension area" in the following step 28 is also "YES". The area information storage area 15 stores the usage status of the extended area, that is, the information indicating the position of the extended area for the protrusion and the information for enlarging the stroke cutout surface (enlarging the recognition limit position). The cutout surface is enlarged. Further, if the handwriting overflows beyond the extended area, step 28 becomes "NO" and an error sound is emitted at step 31.

Thus, when the pen tip is separated from the input tablet 2, the judgment "YES" in step 26 or step 32 is "YES", and in step 33 the input tablet 2
Then, it is checked whether or not the input pen 9 has re-contacted. When the determination of "writing" in step 33 is "YES", it is checked in the next step 34 whether or not the writing position is outside the basic area. If the determination in step 34 is "NO" (the writing position is within the basic area), the process returns to step 25 and the pen position coordinates are written in the coordinate storage area 12. On the other hand, if the determination in step 34 is "YES" (the writing position is outside the basic area), then in step 35 it is determined whether the writing position is within the extended area. It is determined whether or not it is within the enlarged stroke cutout surface. And both steps 3
When both 5 and 36 are "YES", the writing is treated as in the basic area, and the process returns to step 25,
Writing of pen position coordinates is performed.

If, in the above-mentioned series of steps, the determination of "use extended area?" In step 35 is "NO", or the determination of "is it in the enlarged stroke cut-out surface?" In step 36 is "NO", When the input end key of the keyboard is pressed and the judgment of "whether command input is present" in step 37 and the judgment of "input end" in step 38 are both "YES", this input character The process proceeds to the character recognition process after step 39.

First, in step 39, the extended area information storage area 15
Is referred to, and it is determined whether or not the extended area is used. If the judgment is "YES", step 4
After proceeding to 0, the processing such as the reduction illustrated in FIGS. 7 (1) to 7 (4) is executed to fit the handwriting protruding portion into the basic area, and then the character is cut out. And next step 41
In the above, the conversion processing of the 8-direction code is executed for each stroke of the input character, and the converted direction code is stored in the stroke code area 13 of the work area 11.

FIG. 11 shows a code conversion example for the kana character “a”, and in the figure, each direction code of a character stroke is shown. In this code conversion, when adjacent direction codes (in the illustrated example) are connected, as shown in FIG. 12, for example, a rounding process is performed to make the direction code for a long stroke a stroke code.

Next, in step 42, the dictionary collation processing is performed by referring to the dictionary area 17 for the input character pattern including the direction code. When the input character pattern matches any of the standard patterns in the dictionary area 17, the judgment of "matching match" in step 43 becomes "YES" and the process proceeds to step 44, the recognition result is output to the CRT 5, and CR is output.
Characters for recognition are displayed on the T screen. If the determination at step 43 is "NO", proceed to step 45,
An error message is displayed on CRT5, which means that it cannot be recognized.

In step 22 and step 38, when the command input has a content other than the instruction to start or end the tablet input, each judgment is "NO" and the process proceeds to step 46, and the command processing corresponding to the command input is executed. To be executed.

[Brief description of drawings]

FIG. 1 is a perspective view showing the overall configuration of a handwritten character recognition device,
2 and 3 are diagrams for explaining an example of the handwriting protruding from the basic area, FIGS. 4 to 6 are diagrams for explaining an example of setting the extended area, and FIG. 7 (1) to FIG. (4) is a diagram for explaining a processing method for storing the handwriting including the protruding portion in the basic area, FIG. 8 is a diagram showing a memory map in the handwritten character recognition device, and FIG. 9 is for explaining an 8-direction code. FIG. 10, FIG. 10 is a flow chart showing the operation of the apparatus, FIG. 11 is a view for explaining the direction code for the kana character “A”, and FIG. 12 is a view for explaining the rounding process in the code conversion. is there. 1 ... Computer main body 2 ... Input tablet 7 ... Character input area 8 ... Grid 9 ... Input pen A ... Basic area B ... Extended area M ... Stroke cutout surface

Claims (3)

[Claims] 1. An input tablet for detecting contact and separation of a pen-shaped object, and a handwritten character input to the input tablet, in which a grid for writing one character is defined in an area for inputting characters by handwriting. And a recognition processing unit for executing character recognition processing, the recognition processing unit having setting means for setting a predetermined expansion area around a square for pen position detection as a basic area, and the inside of the basic area. Enlargement means for enlarging the area for cutting out one character stroke based on the handwriting and the continuous area from the basic area to the extended area, and the whole handwriting in the cutout area enlarged by the enlarging means is coordinate-converted and stored in the basic area. A handwritten character recognition device comprising: 2. The storage means stores the handwriting in the basic area by reducing the vertical or horizontal length of the entire handwriting in the cut-out area enlarged by the enlarging means. Handwritten character recognition device described. 3. The handwritten character recognizing device according to claim 1, wherein the storage means moves in parallel so that the entire handwriting in the cutout area stored by the enlarging means can be stored in the basic area.